BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//wp-events-plugin.com//6.6.3//EN
TZID:Europe/Paris
X-WR-TIMEZONE:Europe/Paris
BEGIN:VEVENT
UID:1539@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20131121T000000
DTEND;TZID=Europe/Paris:20131121T000000
DTSTAMP:20210605T131857Z
URL:https://iecl.univ-lorraine.fr/events/une-famille-densembles-regenerati
 fs-emboites/
SUMMARY:Une famille d'ensembles régénératifs emboités.
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Philipp
 e Marchal : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1538@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20131128T104500
DTEND;TZID=Europe/Paris:20131128T114500
DTSTAMP:20210605T131857Z
URL:https://iecl.univ-lorraine.fr/events/analyse-du-biais-de-forets-pureme
 nt-aleatoires/
SUMMARY:Analyse du biais de forêts purement aléatoires
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Sylvain
  Arlot : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1537@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20131205T090000
DTEND;TZID=Europe/Paris:20131205T093500
DTSTAMP:20210605T131857Z
URL:https://iecl.univ-lorraine.fr/events/titre-a-preciser-2-2/
SUMMARY:Titre à  préciser
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Anne Br
 iquet : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1552@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20131219T104500
DTEND;TZID=Europe/Paris:20131219T114500
DTSTAMP:20210605T131900Z
URL:https://iecl.univ-lorraine.fr/events/a%c2%89quation-de-tanaka-perturbe
 e-sur-le-graphe-etoile/
SUMMARY:àquation de Tanaka perturbée sur le graphe étoile
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Hatem H
 ajri : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1656@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20140213T104500
DTEND;TZID=Europe/Paris:20140213T110000
DTSTAMP:20210605T132234Z
URL:https://iecl.univ-lorraine.fr/events/larbre-couvrant-minimal-dans-le-p
 lan/
SUMMARY:L'arbre couvrant minimal dans le plan
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Christo
 phe Garban : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1655@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20140220T104500
DTEND;TZID=Europe/Paris:20140220T114500
DTSTAMP:20210605T132218Z
URL:https://iecl.univ-lorraine.fr/events/methode-numerique-pour-le-filtrag
 e-des-systemes-lineaires-a-sauts-markoviens/
SUMMARY:Méthode numérique pour le filtrage des systèmes linéaires à  
 sauts markoviens
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Benoît
 e de Saporta : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1551@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20140227T104500
DTEND;TZID=Europe/Paris:20140227T114500
DTSTAMP:20210605T131900Z
URL:https://iecl.univ-lorraine.fr/events/theoremes-generaux-sur-les-statis
 tiques-dordre-dune-mosaique-aleatoire-et-applications/
SUMMARY:Théorèmes généraux sur les statistiques d'ordre d'une mosaïque
  aléatoire et applications
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Nicolas
  Chenavier : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1550@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20140925T104500
DTEND;TZID=Europe/Paris:20140925T120000
DTSTAMP:20210605T131859Z
URL:https://iecl.univ-lorraine.fr/events/convergence-a-lequilibre-deds-fra
 ctionnaires-multiplicatives/
SUMMARY:Convergence à  l'équilibre d'EDS fractionnaires multiplicatives
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Fabien 
 Panloup : http://www.iecn.u-nancy.fr/~probstat/affseminaires.php
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1549@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20141002T104500
DTEND;TZID=Europe/Paris:20141002T120000
DTSTAMP:20210605T131859Z
URL:https://iecl.univ-lorraine.fr/events/http-www-iecn-u-nancy-fr-probstat
 -affseminaires-php/
SUMMARY:http://www.iecn.u-nancy.fr/~probstat/affseminaires.php
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Daphné
  Dieuleveut : http://www.iecn.u-nancy.fr/~probstat/affseminaires.php
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1540@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20141016T104500
DTEND;TZID=Europe/Paris:20141016T114500
DTSTAMP:20210605T131858Z
URL:https://iecl.univ-lorraine.fr/events/autour-des-marches-aleatoires-lam
 bda-biaisees-sur-un-arbre/
SUMMARY:Autour des marches aléatoires lambda-biaisées sur un arbre
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Loïc d
 e Raphelis : Considérons un arbre de Galton-Watson\, et une marche aléat
 oire aux plus proches voisins sur celui-ci\, avec biais vers le parent. Ap
 rès une description empirique des propriétés simples de cette marche\, 
 nous proposerons dans notre exposé un théorème central limite fonctionn
 el sur la fonction de hauteur du marcheur dans l'arbre. Nous donnerons les
  idées principales de la preuve\, qui se base sur l'étude de la fonction
  de contour de certains arbres bien choisis. Nous verrons aussi que cette 
 preuve s'étend à  l'étude de marches aléatoires sur des arbres avec p
 robabilités de transition elles-même aléatoires.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1536@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20141023T104500
DTEND;TZID=Europe/Paris:20141023T114500
DTSTAMP:20210605T131857Z
URL:https://iecl.univ-lorraine.fr/events/comparaison-de-la-loi-de-deux-ech
 antillons-estimation-adaptive-de-la-densite-relative/
SUMMARY:Comparaison de la loi de deux échantillons : estimation adaptive d
 e la densité relative
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Gaà«l
 le Chagny : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1535@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20141106T104500
DTEND;TZID=Europe/Paris:20141106T114500
DTSTAMP:20210605T131856Z
URL:https://iecl.univ-lorraine.fr/events/integral-approximation-by-kernel-
 smoothing/
SUMMARY:Integral approximation by kernel smoothing
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Franço
 is Portier : https://iecl.univ-lorraine.fr/ProbaStat/affseminaires.php
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1541@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20141120T104500
DTEND;TZID=Europe/Paris:20141120T114500
DTSTAMP:20210605T131858Z
URL:https://iecl.univ-lorraine.fr/events/hypocoercivity-for-degenerate-kol
 mogorov-equations-and-applications-to-the-langevin-dynamics/
SUMMARY:Hypocoercivity for degenerate Kolmogorov equations and applications
  to the Langevin dynamics
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Martin 
 Grothaus : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1542@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20141204T104500
DTEND;TZID=Europe/Paris:20141204T114500
DTSTAMP:20210605T131858Z
URL:https://iecl.univ-lorraine.fr/events/neurones-en-interaction-champ-moy
 en-2/
SUMMARY:Neurones en interaction champ moyen
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Sylvain
  Rubenthaler : On s'intéresse à  un modèle de neurones dit Â«intègr
 e et déchargeÂ». Chaque décharge de neurones influence les autres neur
 ones. On arrive à  une équation de type champ moyen avec un terme d'int
 eraction assez singulier. Les questions sont : l'existence d'une solution 
 en temps long\, la construction d'un système de particules qui converge v
 ers cette équation\, la synchronisation des neurones.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1543@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20150122T104500
DTEND;TZID=Europe/Paris:20150122T114500
DTSTAMP:20210605T131858Z
URL:https://iecl.univ-lorraine.fr/events/controle-stochastique-de-branchem
 ents-diffusions-un-probleme-a-horizon-fini/
SUMMARY:Contrôle stochastique de branchements-diffusions : un problème à
   horizon fini
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Julien 
 Claisse : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1544@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20150409T104500
DTEND;TZID=Europe/Paris:20150409T114500
DTSTAMP:20210605T131858Z
URL:https://iecl.univ-lorraine.fr/events/seminaire-graal-arbres-b-et-urnes
 -de-pa%c2%b3lya/
SUMMARY:[Séminaire GRAAL] Arbres-B et urnes de Pà³lya
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Brigitt
 e Chauvin : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1546@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20150611T104500
DTEND;TZID=Europe/Paris:20150611T114500
DTSTAMP:20210605T131859Z
URL:https://iecl.univ-lorraine.fr/events/marches-au-hasard-sur-des-triangu
 lations-de-delaunay-engendrees-par-des-processus-ponctuels-dans-rd/
SUMMARY:Marches au hasard sur des triangulations de Delaunay engendrées pa
 r des processus ponctuels dans R^d
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Arnaud 
 Rousselle : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1547@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20150910T104500
DTEND;TZID=Europe/Paris:20150910T114500
DTSTAMP:20210605T131859Z
URL:https://iecl.univ-lorraine.fr/events/processus-dhammersley-discrets-et
 -sous-suites-croissantes/
SUMMARY:Processus d'Hammersley discrets et sous-suites croissantes
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Lucas G
 erin : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1545@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20150917T104500
DTEND;TZID=Europe/Paris:20150917T114500
DTSTAMP:20210605T131859Z
URL:https://iecl.univ-lorraine.fr/events/maximum-du-mouvement-brownien-ref
 lechi-sur-un-horizon-fini-et-autres-variables-associees/
SUMMARY:Maximum  du mouvement brownien réfléchi sur un horizon fini et au
 tres variables associées
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Pierre 
 Vallois : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1548@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20150924T104500
DTEND;TZID=Europe/Paris:20150924T114500
DTSTAMP:20210605T131859Z
URL:https://iecl.univ-lorraine.fr/events/processus-de-hawkes-a-memoire-dor
 dre-variable/
SUMMARY:Processus de Hawkes à  mémoire d'ordre variable
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Pierre 
 Hodara : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1526@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20151001T104500
DTEND;TZID=Europe/Paris:20151001T114500
DTSTAMP:20210605T131855Z
URL:https://iecl.univ-lorraine.fr/events/reseau-de-neurones-de-type-integr
 e-et-tire-en-interaction-etude-de-la-synchronisation/
SUMMARY:Réseau de neurones de type intègre et tire en interaction\, étud
 e de la synchronisation
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Etienne
  Tanré : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1525@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20151008T104500
DTEND;TZID=Europe/Paris:20151008T114500
DTSTAMP:20210605T131855Z
URL:https://iecl.univ-lorraine.fr/events/des-flots-rugueux-a-lhomogenisati
 on-des-dynamiques-lentes-rapides/
SUMMARY:Des flots rugueux à  l'homogénisation des dynamiques lentes-rapi
 des
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Rémi C
 atellier : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1524@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20151015T104500
DTEND;TZID=Europe/Paris:20151015T114500
DTSTAMP:20210605T131855Z
URL:https://iecl.univ-lorraine.fr/events/a-chaining-algorithm-for-online-n
 onparametric-regression/
SUMMARY:A Chaining Algorithm for Online Nonparametric Regression
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Sébast
 ien Gerchinovitz : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1523@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20151105T104500
DTEND;TZID=Europe/Paris:20151105T114500
DTSTAMP:20210605T131855Z
URL:https://iecl.univ-lorraine.fr/events/classification-de-signaux-eeg-et-
 filtrage-spatial-par-csp-sparse/
SUMMARY:Classification de signaux EEG et filtrage spatial par CSP sparse
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Marie C
 havent : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1522@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20151112T104500
DTEND;TZID=Europe/Paris:20151112T114500
DTSTAMP:20210605T131854Z
URL:https://iecl.univ-lorraine.fr/events/density-and-tube-estimates-for-di
 ffusions-under-hormander-type-conditions/
SUMMARY:Density and tube estimates for diffusions under Hormander-type cond
 itions
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Paolo P
 igato : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1527@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20151126T104500
DTEND;TZID=Europe/Paris:20151126T114500
DTSTAMP:20210605T131855Z
URL:https://iecl.univ-lorraine.fr/events/comment-est-ce-que-les-bacteries-
 escherichia-coli-croissent/
SUMMARY:Comment est-ce que les bactéries Escherichia coli croissent ?
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Nathali
 e Krell : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1528@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20160107T104500
DTEND;TZID=Europe/Paris:20160107T114500
DTSTAMP:20210605T131855Z
URL:https://iecl.univ-lorraine.fr/events/marche-aleatoire-persistante-recu
 rrence-transience/
SUMMARY:Marche aléatoire persistante : récurrence/transience
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Basile 
 de Loynes : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1534@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20160114T104500
DTEND;TZID=Europe/Paris:20160114T114500
DTSTAMP:20210605T131856Z
URL:https://iecl.univ-lorraine.fr/events/moyennisation-a-la-mode-de-t-g-ku
 rtz-pour-des-processus-deterministes-par-morceaux/
SUMMARY:Moyennisation à  la mode de T. G. Kurtz pour des processus déter
 ministes par morceaux
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Alexand
 re Genadot : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1533@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20160121T104500
DTEND;TZID=Europe/Paris:20160121T114500
DTSTAMP:20210605T131856Z
URL:https://iecl.univ-lorraine.fr/events/estimation-and-clustering-in-a-se
 miparametric-poisson-process-stochastic-block-model-for-longitudinal-netwo
 rks/
SUMMARY:Estimation and clustering in a semiparametric Poisson process stoch
 astic block model for longitudinal networks
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Fanny V
 illers : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1532@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20160128T104500
DTEND;TZID=Europe/Paris:20160128T114500
DTSTAMP:20210605T131856Z
URL:https://iecl.univ-lorraine.fr/events/modelisation-probabiliste-et-infe
 rence-statistique-pour-lanalyse-des-donnees-spatialisees/
SUMMARY:Modélisation probabiliste et inférence statistique pour l'analyse
  des données spatialisées
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Radu St
 oica : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1531@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20160204T104500
DTEND;TZID=Europe/Paris:20160204T114500
DTSTAMP:20210605T131856Z
URL:https://iecl.univ-lorraine.fr/events/tba-2/
SUMMARY:TBA
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Bertran
 d Michel : TBA
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1530@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20160225T104500
DTEND;TZID=Europe/Paris:20160225T114500
DTSTAMP:20210605T131855Z
URL:https://iecl.univ-lorraine.fr/events/bayesian-estimation-by-maximizati
 on-a-posteriori-in-models-with-random-effects-based-on-ordinary-differenti
 al-equations/
SUMMARY:Bayesian estimation by maximization a posteriori in Models with Ran
 dom effects based on Ordinary Differential Equations
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Mélani
 e Prague : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1529@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20160317T104500
DTEND;TZID=Europe/Paris:20160317T114500
DTSTAMP:20210605T131855Z
URL:https://iecl.univ-lorraine.fr/events/testing-uniformity-on-high-dimens
 ional-spheres-against-monotone-rotationally-symmetric-alternatives/
SUMMARY:Testing uniformity on high-dimensional spheres against monotone rot
 ationally symmetric alternatives
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Davy Pa
 indaveine : TBA
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1521@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20161006T104500
DTEND;TZID=Europe/Paris:20161006T114500
DTSTAMP:20210605T131854Z
URL:https://iecl.univ-lorraine.fr/events/abc-shadow-un-outil-pour-lanalyse
 -statistique-des-structures-spatialisees/
SUMMARY:ABC Shadow: un outil pour l'analyse statistique des structures spat
 ialisées
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Radu St
 oica : L'exposé présente l'algorithme ABC Shadow un nouveau algorithme d
 e type ABC (Approximate Bayesian Computation). Cet algorithme peut être u
 tilisé pour échantillonner des lois a posteriori des densités de probab
 ilité qui sont continues et différentiables. Cette méthode parvient a s
 urmonter la principale difficulté de toute méthode ABC\, afin qu'elle pu
 isse àÂªtre utilisée en pratique. Cette difficulté consiste dans le
  fait que l'on a besoin d'un nombre important d'échantillons dans la rég
 ion de l'espace des paramètres\, induite par les statistiques observées.
  L'algorithme est réglé en l'appliquant sur la loi a posteriori d'un mod
 èle Gaussien qui est entièrement connue. Une fois ce réglage effectué\
 , la méthode ABC Shadow est utilisée pour l'analyse des différentes str
 uctures spatialisées. Ces structures proviennent ou bien elles sont suppo
 sées comme des réalisations des processus ponctuels. Les modèles consid
 érés sont : Strauss\, Candy et interaction par aires.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1511@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20161013T104500
DTEND;TZID=Europe/Paris:20161013T114500
DTSTAMP:20210605T131852Z
URL:https://iecl.univ-lorraine.fr/events/flips-contextuels/
SUMMARY:Flips contextuels
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Thomas 
 Fernique : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1510@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20170126T104500
DTEND;TZID=Europe/Paris:20170126T114500
DTSTAMP:20210605T131852Z
URL:https://iecl.univ-lorraine.fr/events/equations-differentielles-stochas
 tiques-avec-temps-local-inhomogenes-en-temps-et-operateurs/
SUMMARY:Equations Différentielles Stochastiques avec temps local inhomogè
 nes en temps\, et opérateurs
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Pierre 
 Etoré : Dans cet exposé on considère des Equations Différentielles Sto
 chastiques (EDS) unidimensionnelles faisant intervenir le temps local du p
 rocessus inconnu\, ainsi que des coefficients discontinus. Ce type d'EDS e
 st en lien avec les opérateurs sous forme divergence à  coefficients di
 scontinus\, ainsi qu'avec les Equations aux Dérivées Partielles (EDP) av
 ec condition de transmission. Ces résultats son assez bien connus dans le
  cas homogène en temps.\nOn se penche ici sur le cas oà¹ tous les coeff
 icients de l'équation dépendent du temps. On montre des résultats d'exi
 stence et d'unicité des solutions pour ce type d'EDS (on étend ainsi des
  résultats pour le cas homogène qui remontent à  J.-F. Le Gall\, 1984)
 . Puis on établit le lien\, via une formule de Feynman-Kac\, entre la sol
 ution de l'EDS et la solution classique d'une EDP parabolique avec conditi
 on de transmission\, et coefficients non-homogènes en temps - en particul
 ier la condition de transmission devient elle-même inhomogène en temps. 
 Nous prouvons nous-mêmes l'existence d'une telle solution classique à  
 l'EDP. Pour ce faire\, on s'appuie sur les travaux de Ladyzhenskaya et al.
  (1966)\, qui ne fournissent toutefois pas le résultat directement. On se
  sert finalement de ces résultats pour étudier le caractère Feller de l
 a solution de l'EDS.\nTravail en commun avec Miguel Martinez de l'UPEMLV.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1509@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20170209T104500
DTEND;TZID=Europe/Paris:20170209T114500
DTSTAMP:20210605T131852Z
URL:https://iecl.univ-lorraine.fr/events/multiple-kernel-learning-applied-
 to-the-integration-of-tara-oceans-datasets/
SUMMARY:Multiple kernel learning applied to the integration of Tara oceans 
 datasets
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Nathali
 e Villa-Vialaneix : The Tara Oceans expedition [sunagawa et al\, 2005] fac
 ilitated the study of plankton communities by providing the scientists wit
 h ocean metagenomic data combined with environmental measures. During the 
 expedition\, 243 seawater samples were collected from 68 locations represe
 nting all main oceanic regions at three depth layers: the surface (SRF)\, 
 the deep chlorophyll maximum (DCM) layer and the mesopelagic (MES) zone. D
 uring the presentation\, I will describe a method to integrate information
  provided by different datasets collected during the expedition. The appro
 ach uses kernels which are combined in an unsupervised setting for data mi
 ning purposes. Additionnaly\, tools to help the interpretation of the resu
 lts are given and shows that well known facts about parts of the datasets 
 are recovered and that new insights on the data are also obtained.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1508@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20170302T104500
DTEND;TZID=Europe/Paris:20170302T114500
DTSTAMP:20210605T131852Z
URL:https://iecl.univ-lorraine.fr/events/percolation-gelee-en-deux-dimensi
 ons/
SUMMARY:Percolation gelée en deux dimensions
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Pierre 
 Nolin : Nous étudions la percolation gelée et les processus de feux de f
 oràÂªts. La percolation gelée est un processus de percolation oà
 Â¹ chaque composante connexe arràÂªte de croàÂ®tre ("gàÂ¨
 le") dàÂ¨s qu'elle devient infinie. Ce modàÂ¨le a été introdui
 t par Aldous en 1999 sur l'arbre binaire\, et nous discutons un processus 
 analogue en deux dimensions\, pour lequel les composantes connexes gàÂ
 ¨lent dàÂ¨s qu'elles contiennent au moins N sommets\, pour un certai
 n paramàÂ¨tre N. En particulier\, nous prouvons que la densité de si
 tes gelés tend vers 0 lorsque N tend vers l'infini\, et nous établissons
  une propriété de "déconcentration" pour les tailles des composantes co
 nnexes. Nous évoquons également des résultats similaires pour les feux 
 de foràÂªts.  ApràÂ¨s une introduction générale sur la percola
 tion indépendante\, nous présentons en détail les principaux outils pou
 r décrire sa transition de phase en deux dimensions. Cette compréhension
  de la percolation pràÂ¨s du point critique joue un ràÂ´le fonda
 mental dans nos résultats.  Cet exposé est basé sur des travaux en coll
 aboration avec Rob van den Berg (CWI and VU\, Amsterdam) et Demeter Kiss (
 U. Cambridge).
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1507@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20170309T104500
DTEND;TZID=Europe/Paris:20170309T114500
DTSTAMP:20210605T131852Z
URL:https://iecl.univ-lorraine.fr/events/le-processus-quantile-rendu-marko
 vien/
SUMMARY:Le processus quantile rendu markovien
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Nicolas
  Juillet : On se donne une famille de mesures réelles indexées sur R. Ce
 s mesures sont les marges de nombreux processus très différents les uns 
 des autres. On pense d'abord au processus indépendant (dont la loi est la
  mesure produit) qui est markovien\, ou au processus quantile (alias comon
 otone) qui préserve l'ordre des quantiles. Je présenterai un nouveau pro
 cessus qui combine les deux propriétés\, dans un sens précis\, et dont 
 je donnerai des exemples variés. Nous verrons une application intéressan
 te à  la théorie du transport et comment l'existence du processus compl
 ète les résultats de Kellerer sur les peacocks. Il s'agit un travail en 
 collaboration avec Charles Boubel.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1506@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20170316T104500
DTEND;TZID=Europe/Paris:20170316T114500
DTSTAMP:20210605T131852Z
URL:https://iecl.univ-lorraine.fr/events/transition-de-phase-pour-le-tasep
 -desordonne/
SUMMARY:Transition de phase pour le TASEP désordonné
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Christo
 phe Bahadoran : Le TASEP est un système de particules sur $ mathbb{Z}$ da
 ns lequel chaque particule essaie indépendamment de sauter d'un pas vers 
 la droite aux instants d'un processus de Poisson\, le saut n'ayant lieu qu
 e si le site visé est libre. C'est un modèle ultra-simplifié de trafic 
 routier. Le flux de particules est une fonction explicite\, strictement co
 ncave\, de la densité moyenne. Ce caractère explicite\, lié à  la con
 naissance explicite des mesures invariantes du processus\, persiste si l'o
 n ajoute du désordre de particules\, en faisant varier l'intensité du pr
 ocessus de Poisson d'une particule à  l'autre (véhicules lents/rapides)
 . On observe alors une transition entre un régime laminaire à  basse de
 nsité\, ou le courant est linéaire\, et un régime d'exclusion à  haut
 e densité. Dans le cas du désordre de site (sites rapides/sites lents)\,
  les mesures invariantes ne sont plus connues\, et la fonction courant n'e
 st plus explicite. Nous démontrons que le courant présente un plateau au
 tour de la densité 1/2 et obtenons la limite explicite de cette fonction 
 et de la taille du plateau dans la limite du désordre dilué. Notre preuv
 e repose sur un argument de renormalisation appliqué à  la percolation 
 de dernier passage\, et des idées d'homogénéisation. \n(Travail en coll
 aboration avec Thierry Bodineau)
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1519@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20170323T091500
DTEND;TZID=Europe/Paris:20170323T101500
DTSTAMP:20210605T131854Z
URL:https://iecl.univ-lorraine.fr/events/sur-le-trou-spectral-des-graphes-
 aleatoires-hyperboliques/
SUMMARY:Sur le trou spectral des graphes aléatoires hyperboliques
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Dieter 
 Mitsche : Le modèle des graphes aléatoires hyperboliques était introdui
 t comme un modèle prometteur pour les réseaux complexes. Nous considéro
 ns le modèle de Krioukov et al. et nous calculons le trou spectral de la 
 Laplacienne de ce modèle. Plus précisément\, nous montrons que $ lambda
 _2$ d'un tel graphe est $ Omega(n^{-(2alpha-1)}/polylog(n))$\, oà¹ $ n$ 
 est le nombre de noeuds et $ 1/2 &lt\; alpha &lt\; 1$ est un paramètre du
  modèle. Nous concluons aussi que la borne supérieure de $ lambda_2$ obt
 enue par l&#039\;inégalité de Cheeger est presque atteinte. Nous caract
 érisons aussi les ensembles des noeuds pour lesquelles cette borne est at
 teinte.\n(travail en collaboration avec Marcos Kiwi)
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1514@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20170323T104500
DTEND;TZID=Europe/Paris:20170323T114500
DTSTAMP:20210605T131853Z
URL:https://iecl.univ-lorraine.fr/events/transition-de-phase-et-proprietes
 -geometriques-de-modeles-de-polymere/
SUMMARY:Transition de phase et propriétés géométriques de modèles de p
 olymère
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Nicolas
  Pétrélis : La Mécanique Statistique s'est révélée être un cadre bi
 en adapté à  l'étudedes propriétés physiques des polymères en inter
 action avec un milieu extérieur.Les questions principales auxquelles les 
 mathématiciens tentent de répondre concernentles transitions de phase (l
 ocalisation\, effondrement etc.) auxquelles les modèles de polymèredonne
 nt lieu\, mais également les propriétés géométriques typiquesd'un pol
 ymère dans un milieu donné\, et à  une température donnée.Dans cet e
 xposé je développerai plus particulièrement les résultats obtenusconce
 rnant la transition d'effondrement d'un modèle de marche aléatoirepartie
 llement dirigée en auto-interaction\, introduit par Zwanzig et Lauritzen 
 (Jour. Chem. Phys. 1968)pour étudier le comportement d'un homopolymère p
 longé dans un solvant pauvre. Je présenterai également un résultatréc
 ent d'existence de la transition d'effondrement pour une variante non diri
 gée de ce modèle.Il s'agit de travaux en collaboration avec Philippe Car
 mona\, Gia Bao Nguyen et Niccolo Torri.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1518@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20170330T091500
DTEND;TZID=Europe/Paris:20170330T101500
DTSTAMP:20210605T131854Z
URL:https://iecl.univ-lorraine.fr/events/stabilite-du-modele-dappariement-
 aleatoire-sur-des-graphes/
SUMMARY:Stabilité du modèle d'appariement aléatoire sur des graphes
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Pascal 
 Moyal : Nous considérons un modèle d'appariement d'entités générées 
 aléatoirement\, pour lequel les paires possibles sont fixées par un grap
 he simple de compatibilité. Ce modèle\, qui a des applications naturelle
 s à  l'économie participative\, la gestion des banques de sang et d'org
 anes et aux chaînes de production\,  généralise celui d'appariement bip
 arti de Kaldentey\, Kaplan et Weiss à  un graphe non-nécessairement bip
 arti. La stabilité du système est étudiée suivant les propriétés str
 ucturelles du graphes de compatibilité. Nous proposons une classe de grap
 hes pour lesquels la zone de stabilité ne dépend pas de la politique d'a
 ppariement  (i.e. l'ordre de priorité en cas de choix multiple)\, et une 
 réciproque partielle. En outre\, nous montrons sous certaines conditions 
 l'existence d'une forme produit particulière pour sa représentation Mark
 ovienne sous la politique "Premier entré\, premier marié." Des connexion
 s de ces résultats avec la théorie classique d'appariement dans les grap
 hes (et dans les grands graphes aléatoires) seront aussi proposées. (tra
 vaux joints avec Jean Mairesse\, Ana Busic et Ohad Perry).
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1512@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20170330T104500
DTEND;TZID=Europe/Paris:20170330T114500
DTSTAMP:20210605T131853Z
URL:https://iecl.univ-lorraine.fr/events/modelisation-de-grands-reseaux-de
 -neurones-par-processus-de-hawkes/
SUMMARY:Modélisation de grands réseaux de neurones par processus de Hawke
 s
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Julien 
 Chevallier : Nous nous intéresserons aux liens qui existent entre deux é
 chelles de modélisation neurobiologique. A un niveau microscopique\, l'ac
 tivité électrique de chaque neurone est représentée par un processus p
 onctuel. à  une plus grande échelle\, un système d'EDP structuré en 
 âge décrit la dynamique moyenne de ces activités. Nous montrerons que l
 e modèle macroscopique (système d'EDP) peut se retrouver à  partir d'u
 n réseau de $ n$ neurones en champ-moyen quand $ n$ tend vers $+infty$ vi
 a une ``Loi des grands nombres''. De plus\, les fluctuations du réseau de
  $ n$ neurones autour du comportement limite/macroscopique sont caractéri
 sées par un ``Théorème central limite''. Cette étude finale permet la 
 dérivation d'un système d'EDP stochastique\, plus proche de la dynamique
  microscopique que le système d'EDP classique.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1517@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20170406T091500
DTEND;TZID=Europe/Paris:20170406T101500
DTSTAMP:20210605T131854Z
URL:https://iecl.univ-lorraine.fr/events/estimation-of-the-realized-co-vol
 atility-vector-large-deviations-approach/
SUMMARY:Estimation of the realized (co-)volatility vector: Large deviations
  approach
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Hacene 
 Djellout : Realized statistics based on high frequency returns have become
  very\npopular in financial economics.In recent years\,\ndifferent non-par
 ametric estimators of the variation of a log-price\nprocess have appeared.
  Among them are the realized\nquadratic (co-)variation which is perhaps th
 e most well known example\,\nproviding a consistent estimator of the integ
 rated\n(co-)volatility when the logarithmic price process is continuous. I
 n\nthis paper\, we propose to study the large deviation\nproperties of rea
 lized (co-)volatility. Our main motivation is to\nimprove upon the existin
 g limit theorems such as the weak law\nof large numbers or the central lim
 it theorem which have been proved in\ndifferent contexts. Our large deviat
 ions results can\nbe used to evaluate and approximate tail probabilities o
 f realized\n(co-)volatility. As an application we provide the large deviat
 ions\nfor the standard dependence measures between the two assets returns 
 such\nas the realized regression coefficients or the\nrealized correlation
 . Our study should contribute to the recent trend of\nresearch on the (co-
 )variance estimation problems\, which\nare quite often discussed in high-f
 requency financial data analysis.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1520@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20170406T104500
DTEND;TZID=Europe/Paris:20170406T114500
DTSTAMP:20210605T131854Z
URL:https://iecl.univ-lorraine.fr/events/some-recent-results-in-optimal-st
 opping/
SUMMARY:Some recent results in optimal stopping
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Ernesto
  Mordecki : After presenting the optimal stopping problem and a classical 
 example\, we present some recent results in optimal stopping for Lévy pro
 cesses and multidimensional diffusions. The possibility of extending this 
 results to one-dimensional diffusion with discontinuous coefficients will 
 be discussed.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1513@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20170427T104500
DTEND;TZID=Europe/Paris:20170427T114500
DTSTAMP:20210605T131853Z
URL:https://iecl.univ-lorraine.fr/events/limite-dechelle-du-processus-de-c
 ontact-sous-critique/
SUMMARY:Limite d'échelle du processus de contact sous-critique
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Aurelia
  Deshayes : Dans cet exposé je parlerai du processus de contact sous crit
 ique. Ce processus modélise une infection qui s'éteint presque sà»reme
 nt si on commence avec un nombre fini de particules infectées et qu'on ch
 oisit un paramètre d'infection suffisamment petit. Mais que se passe-t-il
  si\, dans ce même régime\, on part cette fois d'un nombre infini de par
 ticules infectées? Je présenterai un travail en collaboration avec Leona
 rdo T. Rolla oà¹ nous donnons une description des configurations asympto
 tiques. Une configuration sera décrite par la collection des "emplacement
 s macroscopiques" des zones de particules infectées et par la position re
 lative de ces particules dans chaque zone (faisant intervenir la distribut
 ion quasi stationnaire du processus de contact modulo translation). Ce tra
 vail est une extension d'un résultat de Andjel\, Ezanno\, Groisman et Rol
 la qui décrit le processus de contact sous critique vu depuis la particul
 e la plus à  droite en dimension 1.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1516@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20170504T104500
DTEND;TZID=Europe/Paris:20170504T114500
DTSTAMP:20210605T131854Z
URL:https://iecl.univ-lorraine.fr/events/perpetual-integral-functionals-of
 -brownian-motion/
SUMMARY:Perpetual integral functionals of Brownian motion
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 José A
 lfredo Là³pez Mimbela : We consider semi-linear PDEs perturbed by a mult
 iplicative noise\nof the form\n\n$\n du(t\,x)=[Au(t\,x)+G(u(t\,x))]dt+ku(t
 \,x)dW(t)\,\n$\n\nwhere $ A$ is the Laplacian\, $ G$ is a positive\, incre
 asing convex function\, $ k\n$ is constant and $ {W(t)}$ is a one-dimensio
 nal Brownian motion.\nNontrivial positive solutions of these equations can
  exist globally in time\,\nor they can exhibit blow up in finite time. In 
 this talk we will focus on\nthe later regime and obtain bounds for the blo
 w up times\, which are given in\nterms of integral functionals of $ {W(t)}
 $.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1494@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20170511T104500
DTEND;TZID=Europe/Paris:20170511T114500
DTSTAMP:20210605T131850Z
URL:https://iecl.univ-lorraine.fr/events/a-new-probabilistic-interpretatio
 n-of-keller-segel-model-for-chemotaxis-application-to-1-d/
SUMMARY:A new probabilistic interpretation of Keller-Segel model for chemot
 axis\, application to 1-d
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Milica 
 Tomasevic : The Keller Segel (KS) model for chemotaxis is a two-dimensiona
 l system of parabolic or elliptic PDEs. Motivated by the study of the full
 y parabolic model using probabilistic methods\, we give rise to a non line
 ar SDE of McKean-Vlasov type with a highly non standard and singular inter
 action.\nIn this talk I will briefly introduce the KS model\, point out so
 me of the PDE analysis results on it and then\, in detail\, analyze our pr
 obabilistic interpretation in the case d=1.\nThis is a joint work with D. 
 Talay.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1515@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20170518T104500
DTEND;TZID=Europe/Paris:20170518T114500
DTSTAMP:20210605T131853Z
URL:https://iecl.univ-lorraine.fr/events/concentration-inequalities-for-re
 generative-and-harris-recurrent-markov-chains-with-applications-to-statist
 ical-learning/
SUMMARY:Concentration inequalities for regenerative and Harris recurrent Ma
 rkov chains with applications to statistical learning
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Gabriel
 a Ciolek : Concentration inequalities are very often a crucial step in der
 iving many results in statistical learning. The purpose of this talk is to
  present Bernstein and Hoeffding type maximal inequalities for regenerativ
 e Markov chains. Furthermore\, we generalize these results and show expone
 ntial bounds for suprema of empirical processes over a class of functions 
 F which size is controlled by its uniform entropy number. We show also tha
 t concentration inequalities are possible to obtain when the chain is sub-
 geometric. All constants involved in the bounds of the considered inequali
 ties are given in an explicit form which can be advantageous in practical 
 considerations. We show that the inequalities obtained for regenerative Ma
 rkov chains can be easily generalized to a Harris recurrent case. Finally 
 we provide one example of application of presented inequalities in statist
 ical learning theory and obtain generalization bounds for mimimum volume s
 et estimation problem when the data are Markovian.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1493@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20170615T104500
DTEND;TZID=Europe/Paris:20170615T114500
DTSTAMP:20210605T131850Z
URL:https://iecl.univ-lorraine.fr/events/fonctionnelles-de-coat-sur-des-ar
 bres-aleatoires/
SUMMARY:Fonctionnelles de coà»t sur des arbres aléatoires
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Marion 
 Sciauveau : Les arbres apparaissent naturellement dans de nombreux domaine
 s tels que l'informatique pour le stockage de données ou encore la biolog
 ie pour classer des espèces dans des arbres phylogénétiques.\nDans cet 
 exposé\, nous nous intéresserons aux limites de fonctionnelles additives
  de grands arbres aléatoires. Nous étudierons les cas des arbres binaire
 s sous le modèle de Catalan (arbres aléatoires choisis uniformément par
 mi les arbres binaires enracinés complets ordonnés avec un nombre de nÅ
 “ud donné) et les arbres simplement générés. On obtiendra un princip
 e d'invariance pour ces fonctionnelles ainsi que les fluctuations associé
 es.\nDans le cas binaire\, la preuve repose sur le lien entre  les arbres 
 binaires et l'excursion brownienne normalisée (voir Aldous [1]). Cela nou
 s permettra de retrouver les résultats avancés par Fill et Kapur [2] et 
 Fill et Janson [3].\nRéférences :\n[1] : D. Aldous. The continuum random
  tree. III. (1993)\n[2] : J.A. Fill and N.Kapur. Limiting distributions fo
 r additive functionals on Catalan trees (2004)\n[3] : J.A. Fill and S. Jan
 son. Precise logarithmics for the right tails of some limit random variabl
 es for random trees (2009)
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1492@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20170914T104500
DTEND;TZID=Europe/Paris:20170914T114500
DTSTAMP:20210605T131850Z
URL:https://iecl.univ-lorraine.fr/events/sur-des-equations-aux-derivees-pa
 rtielles-a-coefficients-constants-par-morceaux/
SUMMARY:Sur des équations aux dérivées partielles à  coefficients cons
 tants par morceaux
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Zili Mo
 unir : On présente la solution fondamentale d'une équation aux dérivée
 s partielles à  coefficient constants par morceaux. De telles équations
  apparaissent\, entre autres\, lors de la modélisation de la diffusion de
  particules dans des milieux hétérogènes. Partant d'une représentation
  probabiliste de la solution\, on explicite un développement asymptotique
  en temps petits\, utilisable dans les applications concrètes.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1491@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20170928T104500
DTEND;TZID=Europe/Paris:20170928T114500
DTSTAMP:20210605T131850Z
URL:https://iecl.univ-lorraine.fr/events/time-series-and-long-memory/
SUMMARY:Time series and long memory
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Mariann
 e Clausel : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1495@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20171019T104500
DTEND;TZID=Europe/Paris:20171019T114500
DTSTAMP:20210605T131850Z
URL:https://iecl.univ-lorraine.fr/events/identification-et-caracterisation
 -de-lisotropie-des-champs-aleatoires-deformes-via-leurs-ensembles-dexcursi
 on/
SUMMARY:Identification et caractérisation de l'isotropie des champs aléat
 oires déformés via leurs ensembles d'excursion
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Julie F
 ournier : Une application déterministe $theta$ de $R^2$ dans lui-même d
 éforme le plan de façon bijective et régulière. Avec un champ aléatoi
 re $X$ réel et défini sur $R^2$\, régulier\, stationnaire et isotrope\,
  elle entre dans la construction d'un champ déformé défini comme la com
 posée de $X$ avec $theta$. Un champ déformé est en général anisotrope
 \, cependant certaines applications $theta$\, dont on propose une caracté
 risation explicite\, préservent l'isotropie. En supposant en outre que $X
 $ est gaussien\, on définit une forme faible d'isotropie d'un champ défo
 rmé par une condition d'invariance de la caractéristique d'Euler moyenne
  de certains de ses ensembles d'excursion. On prouve que les champs défor
 més satisfaisant cette définition sont en réalité isotropes en loi. Da
 ns une dernière partie de l'exposé\, en supposant connue la caractérist
 ique d'Euler moyenne de certains ensembles d'excursion d'un champ déform
 é\, on prouve qu'il est possible d'identifier la déformation $theta$ ass
 ociée.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1496@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20171109T104500
DTEND;TZID=Europe/Paris:20171109T114500
DTSTAMP:20210605T131850Z
URL:https://iecl.univ-lorraine.fr/events/risques-garantis-pour-les-systeme
 s-discriminants-multi-classes-a-marge/
SUMMARY:Risques garantis pour les systèmes discriminants multi-classes à
   marge
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Yann Gu
 ermeur : La théorie statistique de l'apprentissage porte sur trois probl
 èmes d'inférence empirique : la discrimination\, la régression et l'est
 imation de la fonction de densité. Cette présentation se concentre sur l
 a discrimination. Nous exposons les garanties disponibles sur les performa
 nces en généralisation des systèmes discriminants (risques garantis)\, 
 en privilégiant le cas oà¹ ceux-ci s'appuient sur le concept de marge. 
 L'intervalle de confiance de ces risques garantis dépend de trois paramè
 tres principaux : la taille m de l'échantillon\, le nombre C de catégori
 es et la valeur gamma du paramètre de marge. Nous caractérisons cette d
 épendance en fonction du choix de la fonction de perte.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1497@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20171116T104500
DTEND;TZID=Europe/Paris:20171116T114500
DTSTAMP:20210605T131850Z
URL:https://iecl.univ-lorraine.fr/events/large-scale-structure-of-the-univ
 erse-observers-point-of-view/
SUMMARY:Large-scale structure of the Universe: observer's point of view
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Elmo Te
 mpel : The cosmic web is a highly complex geometrical pattern\, with galax
 y clusters at the intersection of filaments and filaments at the intersect
 ion of walls. Using observational data\, we can visually recognize the mai
 n components of the cosmic web: voids\, filaments and (super)clusters. How
 ever\, to classify the cosmic web using mathematical methods is much more 
 complicated task\, which also involves the analysis of observational selec
 tion effects. In my talk I will give a brief overview about the observed l
 arge-scale structure together with the main selection effects that should 
 be taken into account while analyzing the data.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1488@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20171123T104500
DTEND;TZID=Europe/Paris:20171123T114500
DTSTAMP:20210605T131849Z
URL:https://iecl.univ-lorraine.fr/events/expose-dans-le-cadre-du-forum-des
 -jeunes-mathematiciennes-et-mathematiciens-lieu-amphi-5-nancy/
SUMMARY:Exposé dans le cadre du Forum des jeunes mathématiciennes et math
 ématiciens (lieu: amphi 5\, Nancy)
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Marie-P
 ierre Etienne : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1505@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20171130T104500
DTEND;TZID=Europe/Paris:20171130T114500
DTSTAMP:20210605T131851Z
URL:https://iecl.univ-lorraine.fr/events/expose-a-la-journee-de-la-federat
 ion-charles-hermite-apprentissage-machine-learning-au-loria/
SUMMARY:Exposé à  la Journée de la Fédération Charles Hermite "Appren
 tissage\, machine learning" au LORIA
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Gilles 
 Blanchard : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1489@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20171207T104500
DTEND;TZID=Europe/Paris:20171207T114500
DTSTAMP:20210605T131849Z
URL:https://iecl.univ-lorraine.fr/events/convergence-de-la-dsf-vers-le-bw/
SUMMARY:Convergence de la DSF vers le BW
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 TRAN Vi
 et Chi : Nous considérons\, sur le plan\, la DSF (Directed Spanning Fores
 t) qui est une forêt dirigée introduite par Baccelli et Bordenave (2007)
 . Soient un processus de Poisson homogène dans le plan et une direction p
 rivilégiée (par exemple -e_y). Nous définissons l'ancêtre de chaque at
 ome du processus de Poisson comme étant l'atome le plus proche (pour la d
 istance euclidienne) et d'ordonnée supérieure. Le graphe résultant est 
 la DSF : il s'agit d'une forêt\, et même presque sà»rement d'un arbre.
  Sous de bonnes renormalisations\, nous montrons que cette forêt converge
  en loi vers la toile Brownienne (BW\, comme Brownian Web). Dans le cas de
  la DSF\, la difficulté majeure est que la construction\, pourtant simple
  et naturelle\, crée des dépendances géométriques très complexes : au
  fur et à  mesure de la construction du graphe\, on accumule une informa
 tion sur la vacuité de certaines régions (aléatoires) du plan. Les crit
 ères de convergence existant dans la littérature s'appuient sur des esti
 mées obtenues en général par la construction de martingales ou chaînes
  de Markov\, constructions qui sont impossibles ici. L'obtention de ces es
 timées clé s'appuie sur des idées de renouvellement fondées sur la gé
 ométrie du problème.\n\nCeci est un travail en commun avec D. Coupier\, 
 K. Saha et A. Sarkar.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1501@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20171214T104500
DTEND;TZID=Europe/Paris:20171214T114500
DTSTAMP:20210605T131851Z
URL:https://iecl.univ-lorraine.fr/events/estimation-of-functional-sparsity
 -in-nonparametric-varying-coefficient-models-for-longitudinal-data-analysi
 s/
SUMMARY:Estimation of Functional Sparsity in Nonparametric Varying Coeffici
 ent Models for Longitudinal Data Analysis
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Juhyun 
 Park : We study the simultaneous domain selection problem for varying coef
 ficient models as a functional regression model for longitudinal data with
  many covariates. The domain selection problem in functional regression mo
 stly appears  under the functional linear regression with scalar response 
 but there is no direct correspondence to functional response models with m
 any covariates. We reformulate the problem as nonparametric function estim
 ation under the notation of "functional sparsity". Sparsity is the recurre
 nt theme that encapsulates interpretability in the face of regression with
  multiple inputs\, and the problem of sparse estimation is well understood
  in the parametric setting as variable selection. For nonparametric models
 \, interpretability not only concerns the number of covariates involved bu
 t also the {em functional form} of the estimates\, and so the sparsity con
 sideration is much more complex. To distinguish the types of sparsity in n
 onparametric models\, we call the former "global sparsity" and the latter 
 "local sparsity"\, which constitute functional sparsity. Most existing met
 hods focus on directly extending the framework of parametric sparsity for 
 linear models to nonparametric function estimation to address one or the o
 ther\, but not both. We develop a penalized estimation procedure that simu
 ltaneously addresses both types of sparsity in a unified framework. We est
 ablish asymptotic properties of estimation consistency and sparsistency of
  the proposed method. Our method is illustrated in simulation study and re
 al data analysis\, and is shown to outperform the existing methods in iden
 tifying both local sparsity and global sparsity.\n\n[this is a joint work 
 with Catherine Y. Tu and Haonan Wang from Colorado State University\, U.S.
 A.]
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1499@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20171221T104500
DTEND;TZID=Europe/Paris:20171221T114500
DTSTAMP:20210605T131851Z
URL:https://iecl.univ-lorraine.fr/events/automates-cellulaires-probabilist
 es-de-memoire-2/
SUMMARY:Automates cellulaires probabilistes de mémoire 2
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Jérôm
 e Casse : Ces travaux ont été effectués en collaboration avec Irène Ma
 rcovici.\n\nUn automate cellulaire probabiliste (ACP) de mémoire 2 est un
  algorithme stochastique qui transforme 2 mots bi-infinis $a = eta_0 = (a_
 i)_{i in Z}$ et $b = eta_1 = (b_i)_{i in Z}$ en un troisième $c = eta_3 =
  (c_i)_{i in Z}$ de tel sorte que la loi de la lettre $c_i$ ne dépend que
  des lettres $(b_i\,a_{i+1}\,b_{i+1})$. Les lettres $c_i$ sont choisies de
  manière synchrone et indépendante. Après avoir obtenu le mot $c$\, on 
 peut ré-appliquer l'ACP en prenant en entrée les mots $(b\,c)$ et ainsi 
 de suite. On obtient alors une suite de mots $(eta_t)_{t &gt\; 0}$ dont le
 s lettres $(eta_t(i))$ forme ce que l'on appelle le diagramme espace-temps
 .\n\nCes ACP de mémoire 2 ont été initialement définis pour étudier l
 e modèle à  8 sommets et nous verrons qu'ils sont également liés à 
  d'autres modèles de la physique statistique comme\, par exemple\, un nou
 veau modèle de TASEP synchrone ou le modèle d'Eden dans le demi-plan sur
  le réseau triangulaire.\n\nDans cet exposé\, nous étudierons les lois 
 invariantes de ces ACP\, l'ergodicité de ces derniers\, ainsi que les pro
 priétés d'invariance de leur diagramme espace-temps. Ce sont des problè
 mes insolubles dans le cas général (y compris pour les ACP à  mémoire
  1) et pour cela nous verrons que nous devons restreindre l'étude à  de
 s cas oà¹ la loi invariante est de type mesure produit ou de type Markov
 .\n\nS'il nous reste un peu de temps à  la fin\, nous verrons comme les 
 méthodes employées dans cet exposé permettent de déduire rapidement de
 s propriétés sur un modèle de TASEP synchrone\ngénéralisé.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1500@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20180111T104500
DTEND;TZID=Europe/Paris:20180111T114500
DTSTAMP:20210605T131851Z
URL:https://iecl.univ-lorraine.fr/events/algorithmes-stochastiques-pour-la
 -statistique-robuste-en-grande-dimension/
SUMMARY:Algorithmes stochastiques pour la statistique robuste en grande dim
 ension
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Antoine
  Godichon : La médiane géométrique est souvent utilisée en statistique
  du fait de sa robustesse. On s'intéresse donc à  des estimateurs rapid
 es de la médiane\, qui consistent en des algorithmes de gradient stochast
 iques moyennés. On définit aussi un nouvel indicateur de dispersion robu
 ste\, appelé Matrice de Covariance Médiane\, avant d'en donner des estim
 ateurs récursifs. Cette matrice\, sous certaines hypothèses\, a les mêm
 es sous-espaces propres que la matrice de covariance\, mais est moins sens
 ible aux données atypiques\, et est donc très intéressante pour l'Analy
 se en Composantes Principales Robuste. Travail joint avec Hervé Cardot et
  Peggy Cénac (Université de Bourgogne).
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1490@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20180118T104500
DTEND;TZID=Europe/Paris:20180118T114500
DTSTAMP:20210605T131850Z
URL:https://iecl.univ-lorraine.fr/events/optimal-best-arm-identification-a
 nd-application-to-monte-carlotree-search/
SUMMARY:(Optimal) Best Arm Identification and application to Monte-Carlo\nT
 ree Search
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Emilie 
 Kaufmann : In Monte-Carlo Tree Search (MCTS)\, the goal is to adaptively\n
 explore paths in a game tree and perform random leaves evaluation\, in\nor
 der to quickly discover the best action to take at the root. In this\ntalk
 \, I will introduce a simple model for MCTS\, that can be viewed as a\nstr
 uctured best arm identification problem in a multi-armed bandit\nmodel. Af
 ter a review of recent advances to tackle the standard best arm\nidentific
 ation (BAI) problem\, I will explain how any BAI algorithm can\nbe convert
 ed to a MCTS algorithm. I will then present empirical results\nand sample 
 complexity guarantees for two particular algorithms\,\nUGapE-MCTS and LUCB
 -MCTS. \n\nThis is joint work with Aurélien Garivier (Université de Toul
 ouse) and\nWouter Koolen (CWI\, Amsterdam)
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1504@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20180125T104500
DTEND;TZID=Europe/Paris:20180125T114500
DTSTAMP:20210605T131851Z
URL:https://iecl.univ-lorraine.fr/events/reconstruction-probabiliste-de-ge
 nealogies-dans-les-populations-vegetales-polyploides/
SUMMARY:Reconstruction probabiliste de généalogies dans les populations v
 égétales polyploïdes
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Frédé
 ric Proia : On proposera dans cet exposé une approche probabiliste de rec
 onstruction de généalogies dans les populations végétales polyploïdes
  (oà¹ les chromosomes ne vont pas nécessairement par paires). On prése
 ntera dans un premier temps une reconstruction dans une population de gen
 êts diploïdes pour lesquels on dispose de la présence/absence de certai
 ns allèles spécifiques : la loi de probabilité du modèle s'appuie sur 
 l'équilibre de Hardy-Weinberg. Dans un second temps\, on généralisera c
 ela à  une population de rosiers polyploïdes\, dont le niveau de ploïd
 ie varie de 2x à  6x (avec une majorité de 4x). Dans un tel modèle\, l
 es lois de reproduction sont soumises à  des règles combinatoires et à
   la problématique du dosage allélique (par exemple un hétérozygote 4
 x peut donner lieu à  de nombreux génotypes : 'aaab'\, 'aabb'\, 'abbb'\
 , 'aabc'\, 'abbc'\, ...\, 'abcd'). Notre modèle tient compte de ces phén
 omènes et propose une arborescence probabilisée des liens génétiques p
 otentiels dans la population.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1503@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20180201T104500
DTEND;TZID=Europe/Paris:20180201T114500
DTSTAMP:20210605T131851Z
URL:https://iecl.univ-lorraine.fr/events/sur-la-contrainte-des-solutions-d
 equations-differentielles-dirigees-par-le-mouvement-brownien-fractionnaire
 /
SUMMARY:Sur la contrainte des solutions d'équations différentielles dirig
 ées par le mouvement brownien fractionnaire.
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Nicolas
  Marie : Dans le contexte du calcul d'Itô\, il existe plusieurs façons d
 e contraindre la solution d'une équations différentielle dirigée par le
  mouvement brownien à  rester dans un convexe fermé de l'espace : probl
 ème de réflexion de Skorokhod\, condition d'invariance\, singularités d
 u champs de vecteurs avec force de rappel etc. Le but de cet exposé est d
 e présenter des extensions de ces méthodes aux équations différentiell
 es dirigées par le mouvement brownien fractionnaire\, dans le contexte de
  la théorie des trajectoires rugueuses.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1487@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20180208T104500
DTEND;TZID=Europe/Paris:20180208T114500
DTSTAMP:20210605T131849Z
URL:https://iecl.univ-lorraine.fr/events/economic-and-financial-problemati
 c-in-discrete-time-models-with-multiple-and-non-dominated-priors/
SUMMARY:Economic and financial problematic in  discrete-time models with mu
 ltiple and non-dominated priors
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Laurenc
 e Carassus : We will present some financial and economic problematics aris
 ing in discrete-time financial/economic models with a finite time horizon 
 under non-dominated model uncertainty. This means that there exists a set 
 of probability measures representing the agent beliefs and that this set i
 s not dominated by a reference measure.\n\nThe technics are based on dynam
 ic programming and measurable selection. We also use analytic sets which d
 isplay the nice property of being stable by projection or countable unions
  and intersections but fail to be stable by complementation.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1502@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20180215T104500
DTEND;TZID=Europe/Paris:20180215T114500
DTSTAMP:20210605T131851Z
URL:https://iecl.univ-lorraine.fr/events/variational-inference-in-the-pois
 son-lognormal-model-for-multivariate-analysis-in-ecology/
SUMMARY:Variational Inference in the Poisson lognormal model for multivaria
 te analysis in ecology
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Julien 
 Chiquet : Many application domains such as ecology or genomics have to dea
 l with multivariate count data. A typical example is the joint observation
  of the respective abundances of a set of species in a series of sites\, a
 iming to understand the co-variations between these species. The Gaussian 
 setting provides a canonical way to model such dependencies\, but does not
  apply in general. We adopt here the Poisson lognormal (PLN) model\, which
  is attractive since it allows one to describe multivariate count data wit
 h a Poisson distribution as the emission law\, while all the dependencies 
 is kept in an hidden friendly multivariate Gaussian layer. While usual max
 imum likelihood based inference raises some issues in PLN\, we show how to
  circumvent this issue by means of a variational algorithm for which gradi
 ent descent easily applies. We then derive several variants of our algorit
 hm to apply PLN to PCA\, LDA and sparse covariance inference on multivaria
 te count data. We illustrate our method on microbial ecology datasets\, an
 d show the importance of accounting for covariate effects to better unders
 tand interactions between species.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1498@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20180222T104500
DTEND;TZID=Europe/Paris:20180222T114500
DTSTAMP:20210605T131851Z
URL:https://iecl.univ-lorraine.fr/events/mean-field-games-with-branching/
SUMMARY:Mean-field games with branching
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Xiaolu 
 Tan : The mean-field game (MFG) consists in a differential game of a very 
 large population on a finite or infinite time horizon. In this work\, we s
 tudy the mean-field games with branching\, which allows to model the immig
 ration\, default and reproduction behaviour in the population. We show how
  the branching feature would change the formulation of the problem and the
 n provide a general existence result of this new MFG. This is a joint work
  with Julien Claisse and Zhenjie Ren
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1474@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20180322T104500
DTEND;TZID=Europe/Paris:20180322T114500
DTSTAMP:20210605T131848Z
URL:https://iecl.univ-lorraine.fr/events/percolation-arithmetique/
SUMMARY:Percolation arithmétique
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Sébast
 ien Martineau : Si on part du réseau carré et efface chaque sommet indé
 pendamment avec une certaine probabilité q\, on effectue ce qui s'appelle
  une percolation de Bernoulli : cet important modèle de mécanique statis
 tique rend compte des phénomènes d'infiltration en milieu poreux. Si on 
 part du réseau carré mais cette fois-ci efface chaque sommet $(x\,y)$ te
 l que $PGCD(x\,y)neq 1$\, on obtient maintenant un objet déterministe de 
 nature arithmétique. Est-il possible de former une percolation (véritabl
 ement aléatoire donc) riche en informations arithmétiques ?\n\nOn va voi
 r que cela est effectivement possible : on peut définir à  quoi ressemb
 le le sous-graphe arithmétique précédent "vu depuis un point tiré unif
 ormément dans le plan". Ce sous-graphe aléatoire est obtenu selon un "cr
 ible d'Ératosthène aléatoire". On fournira de ce graphe aléatoire une 
 définition élémentaire\, puis utilisera le lemme chinois pour faire le 
 pont entre le sous-graphe arithmétique déterministe et sa contrepartie a
 léatoire. On abordera ensuite brièvement quelques problèmes naturels\, 
 comme l'étude des composantes connexes infinies du graphe aléatoire.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1475@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20180405T104500
DTEND;TZID=Europe/Paris:20180405T114500
DTSTAMP:20210605T131848Z
URL:https://iecl.univ-lorraine.fr/events/comportement-asymptotique-des-mat
 rices-aleatoires-de-wishart-gaussiennes-correlees-en-grande-dimension/
SUMMARY:Comportement asymptotique des matrices aléatoires de Wishart gauss
 iennes corrélées en grande dimension
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Ivan No
 urdin : Nous considérerons des matrices de Wishart en grande dimension\, 
 dont les coefficients sont des gaussiennes possiblement corrélées. Dans 
 la situation "mémoire courte"\, nous analyserons la proximité en loi de 
 ces matrices avec l'ensemble gaussien correspondant quand la taille de la 
 matrice tend vers l'infini\, au sens de la distance de Wasserstein. Dans l
 a situation "mémoire longue"\, la situation est tout autre: nous mettrons
  en évidence la convergence vers une matrice aléatoire\, que nous avons 
 appelée matrice de Rosenblatt-Wishart. Cet exposé sera basé sur un trav
 ail en collaboration avec Guangqu Zheng (Univ. Luxembourg).
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1476@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20180412T104500
DTEND;TZID=Europe/Paris:20180412T114500
DTSTAMP:20210605T131848Z
URL:https://iecl.univ-lorraine.fr/events/modele-alpha-cir-et-ses-applicati
 ons-en-finance/
SUMMARY:Modèle alpha-CIR et ses applications en finance
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Simone 
 Scotti : Les processus de branchement avec immigration (CBI) est une class
 e bien étudiée en probabilité. En particulier\, les travaux de Dawson e
 t Li ont fourni une écriture\nexplicite d'une EDS pour un CBI général e
 t prouvé l'unicité de sa solution.\nNous exploitons cette écriture pour
  proposer une extension du modèle CIR pour inclure une partie à  sauts 
 dans le mécanisme de branchement. Nous considérons en particulier le cas
  alpha-stable à  cause de sa parcimonie et pour sa cohérence avec des r
 ésultats statistiques.\nNous montrons que ce modèle\, appliqué aux taux
  d'intérêt\, permet d'expliquer plusieurs effets\nconnus sur les marché
 s obligataires comme la persistance des taux faibles malgré la présence 
 des grands sauts.\nUne deuxième application est l'extension du modèle He
 ston pour inclure des sauts auto-excités dans le processus variance. Nous
  étudions en particulier le comportement de la volatilité implicite sur 
 l'action et sur ses variance swap. Nous testons ce modèle sur les donnes 
 du VIX en montrant que notre modèle fit bien.\nBasé sur deux travaux ave
 c Ying Jiao\, Chunhua Ma et Chao Zhou.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1477@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20180419T104500
DTEND;TZID=Europe/Paris:20180419T114500
DTSTAMP:20210605T131848Z
URL:https://iecl.univ-lorraine.fr/events/unbiased-simulation-methods-based
 -on-the-parametrix-ii/
SUMMARY:Unbiased simulation methods based on the parametrix II
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Arturo 
 Kohatsu-Higa : In these two presentations\, we will first introduce using 
 basic stochastic calculus\, the parametrx method and then show how to dedu
 ce an unbiased simulation method and its interpretations. \n\nWe will disc
 uss its advantages and shortcomings and then discuss how to solve them. \n
 using a second\norder method. \nWe will also give some simulation results 
 and then time allowing we will discuss some other extensions.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1479@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20180531T104500
DTEND;TZID=Europe/Paris:20180531T114500
DTSTAMP:20210605T131848Z
URL:https://iecl.univ-lorraine.fr/events/sur-un-algorithme-dexploration-po
 ur-des-jeux/
SUMMARY:Sur un algorithme d'exploration pour des jeux
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Nicolas
  Fournier : On essaye d'étudier une classe d'algorithmes du type Monte Ca
 rlo Tree Search pour des jeux\ndéterministes à  trait alterné et infor
 mation complète (du type morpion\, puissance 4\, etc.).
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1486@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20180614T104500
DTEND;TZID=Europe/Paris:20180614T114500
DTSTAMP:20210605T131849Z
URL:https://iecl.univ-lorraine.fr/events/processus-auto-repulsifs-et-metad
 ynamique/
SUMMARY:Processus auto-répulsifs et métadynamique
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Pierre-
 André Zitt : Un défaut habituel des algorithmes de Monte Carlo par chaî
 nes de Markov\nest leur difficulté à  explorer l'espace convenablement\
 , ce qui mène\nà  de grandes erreurs d'estimation. L'algorithme de "mé
 tadynamique"\,\nintroduit par Bussi\, Laio et Parrinello dans les années 
 2000 illustre l'une des stratégies possibles pour contourner la difficult
 é\, en gardant la mémoire de la trajectoire passée et en l'utilisant po
 ur biaiser le processus et le pousser vers des régions peu visitées.\nL'
 analyse des processus sous-jacent n'est pas aisée en général \; nous di
 scuterons d'un modèle jouet\, que l'on peut traiter par des outils de la 
 littérature des processus auto-répulsifs.\n\nTravail en commun avec B. J
 ourdain et T. Lelièvre (Ecole des Ponts ParisTech).
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1480@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20180621T104500
DTEND;TZID=Europe/Paris:20180621T114500
DTSTAMP:20210605T131848Z
URL:https://iecl.univ-lorraine.fr/events/extinction-de-populations-faiblem
 ent-inhomogenes-en-dimension-deux/
SUMMARY:Extinction de populations faiblement inhomogènes en dimension deux
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Kilian 
 Raschel : Les populations composées de deux types d'individus (ou phénot
 ypes) peuvent naturellement s'étudier grâce aux marches aléatoires plan
 aires. Dans cet exposé je ferai un survol des techniques existantes et pr
 ésenterai également de nouvelles idées pour l'étude de telles populati
 ons bidimensionnelles. Nous commencerons par le cas des populations homog
 ènes\, qui correspondent à  des modèles maintenant classiques de march
 es aléatoires dans le quart de plan. Nous présenterons ensuite une class
 e de marches aléatoires inhomogènes ayant une interprétation biologique
  naturelle. En passant nous lierons ces marches aléatoires inhomogènes a
 vec d'autres processus\, provenant de la théorie des files d'attente (joi
 ndre la file la plus courte)\, de processus de branchement\, de modèles d
 'urnes et de théorie du potentiel. Il s'agit de différents travaux en co
 mmun avec Gerold Alsmeyer (Mà¼nster)\, Irina Kurkova (Paris 6)\, Pauline
  Lafitte (ECP) et Chi Tran (Lille).
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1481@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20180628T104500
DTEND;TZID=Europe/Paris:20180628T114500
DTSTAMP:20210605T131849Z
URL:https://iecl.univ-lorraine.fr/events/on-shrinkage-estimators-in-multiv
 ariate-models-with-change-points/
SUMMARY:On Shrinkage estimators in multivariate models with change-points
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Sévér
 ien Nkurunziza : In  this  talk\, we  present  some  inference  methods  i
 n  some multivariate models with\nmultiple unknown change-points when the 
 target parameter is suspected to satisfy an uncertain constraint. We waive
  the assumptions on the error terms and establish the joint asymptotic nor
 mality of the unrestricted estimator and the restricted estimator. Further
 \, we propose a class of shrinkage estimators that includes as a special c
 ase the unrestricted estimator\, the estimator restricted as well as James
 -Stein type estimators. To study the performance of the proposed estimator
 s\, we generalize some classical identities underlying the multivariate Ga
 ussian random samples or\, more generally\, the multivariate elliptically 
 contoured random samples. Finally\, we prove that shrinkage estimators dom
 inate the unrestricted estimator.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1485@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20180920T104500
DTEND;TZID=Europe/Paris:20180920T114500
DTSTAMP:20210605T131849Z
URL:https://iecl.univ-lorraine.fr/events/oscillations-pour-des-systemes-de
 -neurones-en-interactions-modelises-par-des-processus-de-hawkes-non-lineai
 res/
SUMMARY:Oscillations pour des systèmes de neurones en interactions modéli
 sés par des processus de Hawkes non-linéaires
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Eva Lö
 cherbach : Nous considérons des systèmes de processus de Hawkes structur
 és en un nombre fini de populations\, avec des interactions du type champ
  moyen. Ces processus décrivent les instants de décharge électrique de 
 neurones en interactions. Nous décrivons la limite lorsque le nombre de n
 eurones par population tend vers l'infini et nous montrons que dans certai
 ns cas\, le processus limite possède des solutions périodiques qui corre
 spondent aux orbites périodiques d'un système dynamique associé au proc
 essus limite. Nous établissons aussi des théorèmes limites qui décrive
 nt la convergence du système de neurones à  taille finie vers ces orbit
 es périodiques. \n\nCes résultats sont basés sur un travail en collabor
 ation avec S. Ditlevsen.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1484@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20180927T104500
DTEND;TZID=Europe/Paris:20180927T114500
DTSTAMP:20210605T131849Z
URL:https://iecl.univ-lorraine.fr/events/structured-cell-population-dynami
 cs-applied-to-the-early-development-of-ovarian-follicles/
SUMMARY:Structured cell population dynamics applied to the early developmen
 t of ovarian follicles
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Frédé
 rique Robin : The ovarian follicles are the basic anatomical and functiona
 l units of the ovaries\, which are renewed from a\nquiescent pool all alon
 g reproductive life.  Follicular development involves a finely tuned seque
 nce of growth and\nmaturation processes\, involving complex cell dynamics.
   In their early stages of development\, ovarian follicles are made up of 
 a germ cell (oocyte)\, whose diameter increases steadily\, and of surround
 ing proliferating somatic\ncells\, which are layered in a globally spheric
 al and compact structure.\nHere\, we present two complementary modeling ap
 proaches dedicated to the first stages of a follicle develop-\nment\, star
 ting with the exit from the pool of quiescent (primordial) follicles leadi
 ng to growth initiation\, and\nending up just before the breaking of the s
 pherical symmetry induced by the follicle cavitation (formation of\nthe an
 trum cavity).\nThe initiation phase is described by joint stochastic dynam
 ics accounting for cell shape transitions (from\na  flattened  to  a  cubo
 idal  shape)  and  proliferation  of  reshaped  cells.   We  can  derive  
 the  mean  time  elapsed before all cells have changed shapes and the corr
 esponding increment in the total cell number\, which is fitted\nto  experi
 mental  data  retrieved  from  primordial  follicles  (single  layered  fo
 llicle  with  only  flattened  cells)  and primary follicles (single layer
 ed follicles with only cuboidal cells).\nThe  next  stages\,  characterize
 d  by  the  accumulation  of  cell  layers  around  the  oocyte\,  are  de
 scribed  by\nmulti-type  structured  models  in  either  a  stochastic  or
   deterministic  framework.   We  have  designed  a  linear age-structured
  stochastic (Bellman-Harris branching) process ruling the changes in the n
 umber of follicular cells and their distribution into successive layers\, 
 which is inspired from the nonlinear model initially introduced in [1]\, a
 s well as is deterministic counterpart (multi-dimensional Mc Kendrick Von 
 Foerster).  We have studied the large-time behavior of the models and deri
 ved explicit analytical formulas characterizing an exponential growth\nof 
 the population (Malthus parameter\, asymptotic cell number moments and sta
 ble age distribution). We have\ncompared the theoretical and numerical out
 puts of the models with experimental biological data informing on follicle
   morphology  in  the  ovine  species  (follicle  and  oocyte  diameters\,
   layer  number  and  total  cell  number) from the primary to the pre-ant
 ral stage.  In addition\, in the case of age independent division rates\, 
 we have established the structural identifiability of the parameters\, and
  estimated the parameter values fitting the cell numbers in each layer dur
 ing the early stages of follicle development.\n\n[1  ]  Clément  F.\,  Mi
 chel  P.\,  Monniaux  D.\,  Stiehl  T.\,  Coupled  somatic  cell  kinetics
   and  germ  cell  growth:\nmutliscale model-based insight on ovarian foll
 icular development\,\nMultiscale  et  Modeling  &amp\;  Simulation\n\, 11(
 3)\, 719-746\, 2013.\n\n[2  ]  Clément F.\,  Robin F.\,  Yvinec R.\,  Ana
 lysis and  calibration of  a linear  model for  structured cell  populatio
 ns  with  unidirectional  motion  :  Application  to  the  morphogenesis  
 of  ovarian  follicles\,\nSubmitted. https://arxiv.or/abs/1712.05372
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1463@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20181004T104500
DTEND;TZID=Europe/Paris:20181004T114500
DTSTAMP:20210605T131846Z
URL:https://iecl.univ-lorraine.fr/events/bayesian-inference-approach-of-me
 chanistic-models-to-date-and-localize-an-invasion/
SUMMARY:Bayesian inference approach of mechanistic models to date and local
 ize an invasion
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Candy A
 bboud : Population dynamics of pathogens invading new territories continue
 s to be of primary concern for both biologists and mathematicians. Extensi
 ve researches are mainly carried out throughout mathematical modeling to r
 econstruct the past dynamics of the alien species.We present a mechanistic
 -statistical approach that allows us to date and localize the invasion of 
 an alien species and describe other epidemiological parameters as per exam
 ple\, the diffusion\, the reproduction and the mortality parameter. The us
 ed approach is based on (i) a coupled reaction-diffusion-absorption sub-mo
 del that describes the dynamics of the epidemics in a heterogeneous domain
  and (ii) a stochastic sub-model that represents the observation process. 
 Then\, we will jointly estimate the initial conditions (date and site) and
  the epidemiological parameters using a Bayesian framework through an adap
 tive multiple importance sampling algorithm. We will show the results obta
 ined in this framework on the basis of abundant post-introduction data gat
 hered to draw up a surveillance plan on the expansion of Xylella fastidios
 a\, a phytopathogenic bacterium detected in South Corsica in 2015. Neverth
 eless\, this approach could be applied to other post-emergent species in o
 rder to endorse a fast reaction.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1462@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20181011T104500
DTEND;TZID=Europe/Paris:20181011T114500
DTSTAMP:20210605T131846Z
URL:https://iecl.univ-lorraine.fr/events/bump-detection-in-gaussian-observ
 ations/
SUMMARY:Bump detection in Gaussian observations
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Farida 
 enikeeva : I will talk about the problem of detection of a change in mean 
 in a Gaussian vector. A bump is a stepwise change within an interval of a 
 given length but unknown location. We consider the problem of heterogeneou
 s bump detection when the change occurs in mean and in variance of the obs
 erved vector and the detection of a bump for dependent observations. Minim
 ax detectability conditions will be presented. \nJoint work with A. Munk\,
  M. Pohlmann and F. Werner.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1482@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20181018T104500
DTEND;TZID=Europe/Paris:20181018T114500
DTSTAMP:20210605T131849Z
URL:https://iecl.univ-lorraine.fr/events/algorithmes-dexploration-markovie
 ns-sur-les-grands-graphes-aleatoires-et-applications/
SUMMARY:Algorithmes d'exploration Markoviens sur les grands graphes aléato
 ires et applications
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Pascal 
 Moyal : Nous présentons plusieurs algorithmes d'exploration de graphes al
 éatoires\, markoviens dans le sens oà¹ leur implémentation est simulta
 née à  la construction-même du graphe par le modèle de configuration.
  Pour différents modèles\, par des approximations fluides des processus 
 markoviens sous-jacents\, nous obtenons des estimations en grand graphe de
  (i) la taille de la famille indépendante maximale\, avec des application
 s au protocole de télécommunication CSMA\; (ii) la dynamique d'une épid
 émie de type SIR sur un réseau hétérogène et (iii) la taille d'un cou
 plage maximal sur un grand graphe aléatoire\, éventuellement orienté.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1483@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20181108T104500
DTEND;TZID=Europe/Paris:20181108T114500
DTSTAMP:20210605T131849Z
URL:https://iecl.univ-lorraine.fr/events/contributions-in-quantitative-mod
 eling-of-metastasis/
SUMMARY:Contributions in quantitative modeling of metastasis
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Sébast
 ien Benzekry : In the majority of solid cancers\, secondary tumors (metast
 ases) and associated complications are the main cause of death. In order t
 o define the optimal therapeutic strategy for a given patient\, one of the
  major current challenges is to estimate\, at diagnosis\, the burden of in
 visible metastases and how they will respond to treatments. In this talk\,
  I will present research efforts towards the establishment of a predictive
  computational tool of metastatic development\, with a particular emphasis
  on the assessment of mathematical models to empirical data (both experime
 ntal and clinical). I will first present the model's framework\, which is 
 based on a physiologically-structured partial differential equation for th
 e time dynamics of a population of metastases\, combined to a nonlinear mi
 xed-effects model for statistical representation of the distribution of th
 e parameters in the population. Then\, I will show results about the descr
 iptive power of the model on data from clinically relevant ortho-surgical 
 animal models of metastasis (breast and kidney tumors)\, with recent findi
 ngs about differential effect of therapies between primary and secondary t
 umors. The talk will further be devoted to the translation of this modelin
 g approach toward the clinical reality. Using clinical imaging data of bra
 in metastasis from non-small cell lung cancer\, several biological process
 es will be investigated to establish a minimal and biologically realistic 
 model able to describe the data. Integration of this model into a biostati
 stical approach for individualized prediction of the model's parameters fr
 om data only available at diagnosis will also be discussed. Together\, the
 se results represent a step forward towards the integration of mathematica
 l modeling as a predictive tool for personalized medicine in oncology.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1478@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20181115T104500
DTEND;TZID=Europe/Paris:20181115T114500
DTSTAMP:20210605T131848Z
URL:https://iecl.univ-lorraine.fr/events/fleming-viot-particle-systems-asy
 mptotic-behavior-and-illustration-inmolecular-dynamics/
SUMMARY:Fleming-Viot particle systems: asymptotic behavior and illustration
  in\nmolecular dynamics
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Arnaud 
 Guyader : The distribution of a Markov process with killing\, conditioned 
 to be\nstill alive at a given time\, can be approximated by a Fleming-Viot
 \nparticle system. In such a system\, each particle is simulated\nindepend
 ently according to the law of the underlying Markov process\, and\nbranche
 s onto another particle at each killing time. The purpose of this\ntalk is
  to present a central limit theorem for the law of the\nFleming-Viot parti
 cle system at a given time in the large population\nlimit. We will illustr
 ate this result on an application in molecular\ndynamics. This is a joint 
 work with Frédéric Cérou\, Bernard Delyon and\nMathias Rousset.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1465@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20181122T104500
DTEND;TZID=Europe/Paris:20181122T114500
DTSTAMP:20210605T131846Z
URL:https://iecl.univ-lorraine.fr/events/quasicrystal-phases-in-a-finite-r
 ange-lattice-gas-model/
SUMMARY:Quasicrystal phases in a finite-range lattice gas model
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Siamak 
 Taati : In a quasicrystal\, the arrangement of the atoms is highly ordered
  (as\nin an ordinary crystal) but non-periodic (unlike in a crystal).  The
 re\nare various mathematical challenges in connection with quasicrystals.\
 nFrom the point of view of statistical mechanics\, the major open\nproblem
  is to provide a mathematical explanation of the formation and\nstability 
 of quasicrystals in presence of thermal fluctuations.  In\nthis talk\, I w
 ill present a (toy) lattice gas model with finite-range\ninteractions that
  has stable quasicrystal phases at positive\ntemperature (i.e.\, Gibbs mea
 sures supported at perturbations of\nnon-periodic tilings).  The construct
 ion is based on old results on\ncellular automata and tilings\, in particu
 lar\, a method of simulating\none cellular automaton with another that is 
 resilient against noise\,\nand the existence of aperiodic sets of Wang til
 es that are\ndeterministic in one direction.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1466@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20181129T104500
DTEND;TZID=Europe/Paris:20181129T114500
DTSTAMP:20210605T131846Z
URL:https://iecl.univ-lorraine.fr/events/an-open-problem-in-ruin-theory-an
 d-its-diffusion-approximation-regime/
SUMMARY:An open problem in ruin theory and its diffusion approximation regi
 me
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Nabil K
 azi-Tani : The De Vylder and Goovaerts conjecture is an open problem in ri
 sk theory\, stating that the finite time ruin probability in a standard ri
 sk model is greater or equal to the corresponding ruin probability evaluat
 ed in the associated model with equalized claim amounts. Equalized means h
 ere that the jump sizes of the associated model are equal to the average j
 ump in the initial model between 0 and a terminal time T.\nIn this talk\, 
 we will consider the diffusion approximations of both the standard risk mo
 del and the associated risk model. We will prove that the associated model
 \, when conveniently renormalized\, converges in distribution to a gaussia
 n process satisfying a simple SDE with explicit coefficients. We will then
  compute the probability that this diffusion hits the level 0 before time 
 T and compare it with the same probability for the diffusion approximation
  for the standard risk model\, which is well known. We will then conclude 
 that the De Vylder and Goovaerts conjecture holds true for these diffusion
  limits.\nThis is a joint work with Stefan Ankirchner (University of Jena)
  and Christophette Blanchet-Scalliet (Ecole Centrale de Lyon and ICJ).
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1458@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20181206T104500
DTEND;TZID=Europe/Paris:20181206T114500
DTSTAMP:20210605T131845Z
URL:https://iecl.univ-lorraine.fr/events/amaretto-multi-omics-data-fusion-
 for-cancer-data/
SUMMARY:AMARETTO: Multi-omics data fusion for cancer data
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Magali 
 Champion : Integrating the increasing number of available multi-omics canc
 er data remains one of the main challenges to improve our understanding of
  cancer. Our approach is based on AMARETTO\, an algorithm that integrates 
 DNA methylation\, DNA copy number and gene expression data to identify can
 cer driver genes and associates them to modules of co-expressed genes. We 
 then propose a pancancer version of AMARETTO by connecting all modules in 
 pancancer communities. This leads to the identification of major oncogenic
  pathways and master regulators involved in different cancers.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1459@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20181213T104500
DTEND;TZID=Europe/Paris:20181213T114500
DTSTAMP:20210605T131846Z
URL:https://iecl.univ-lorraine.fr/events/theorie-des-perturbations-basee-s
 ur-une-nouvelle-formule-dintegration-par-parties-non-lineaire/
SUMMARY:Théorie des perturbations basée sur une nouvelle formule d'intég
 ration par parties non linéaire
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Sara Ma
 zzonetto : Pendant le séminaire\, nous introduirons une formule d'intégr
 ation par partie non linéaire qui peut être vu comme une généralisatio
 n stochastique du lemme de Alekseev-Gröbner.\n\nLa preuve est basée sur 
 le calcule de Malliavin et sur l'expression de certains intégrales stocha
 stiques anticipatifs comme intégrales de Skorohod.\n\nLa formule que l'on
  présente induit une théorie de perturbations\, i.e. une façon d'estime
 r\, en terme de caractéristiques locales\, l'erreur globale entre la solu
 tion exacte d'une équation différentielle stochastique et un processus d
 'Itô quelconque.\n\nSi le temps le permet\, nous parlerons des différenc
 es par rapport au résultat de perturbation établi précédemment par M. 
 Hutzenthaler et A. Jentzen\, et des applications comme la dérivation des 
 taux de convergence en moyenne quadratique des schémas d'approximations p
 our ED(P)S.\n\n(Travail en collaboration avec A. Hudde\, M. Hutzenthaler\,
  et A. Jentzen)
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1457@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20190110T104500
DTEND;TZID=Europe/Paris:20190110T114500
DTSTAMP:20210605T131845Z
URL:https://iecl.univ-lorraine.fr/events/a-polarization-oriented-framework
 -for-bivariate-random-signals/
SUMMARY:A polarization-oriented framework for bivariate random signals
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Julien 
 Flamant : Bivariate signals appear in a broad range of applications: polar
 ized waveforms in seismology and optics\, current velocities in oceanograp
 hy\, etc. Formally\, bivariate signals are 2D vector time series. Existing
  approaches for bivariate signal processing do not provide a straightforwa
 rd description of the signal in terms of its polarization properties. For 
 this purpose we introduce a new and generic framework based on a tailored 
 quaternion Fourier transform. \nThis new framework re-establishes a clear 
 interpretability in terms of polarization attributes of usual quantities s
 uch as spectral densities\, linear filters\, etc. \nIn this talk\, I will 
 introduce the main features of this approach\, with the focus on second-or
 der stationary random bivariate signals. I will discuss spectral analysis\
 , linear filtering and some original decompositions of bivariate signals. 
 Synthetic data will illustrate the usefulness of the proposed framework.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1464@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20190117T104500
DTEND;TZID=Europe/Paris:20190117T114500
DTSTAMP:20210605T131846Z
URL:https://iecl.univ-lorraine.fr/events/homogeneisation-pour-les-mouvemen
 ts-cinetiques-ergodiques/
SUMMARY:Homogénéisation pour les mouvements cinétiques ergodiques
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Pierre 
 Perruchaud : Un exemple de mouvement cinétique est celui d'une particule\
 , soumise à \ndes chocs aléatoires. En supposant que les chocs encodent
 \nl'accélération\, la vitesse suit une équation différentielle\nstocha
 stique\, tandis que la position intègre simplement la vitesse. Le\nmouvem
 ent résultant peut être assez délicat à  étudier\, si par exemple la
 \nparticule est contrainte à  rester sur une surface\, ou que la dynamiq
 ue\nde la vitesse est complexe. Je montrerai que sous des hypothèses trè
 s\nsimples de symétrie et d'ergodicité pour le processus vitesse\, le\np
 rocessus convenablement renormalisé converge vers un mouvement brownien\n
 lorsque les chocs augmentent en intensité.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1461@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20190124T104500
DTEND;TZID=Europe/Paris:20190124T114500
DTSTAMP:20210605T131846Z
URL:https://iecl.univ-lorraine.fr/events/do-we-need-a-new-cosmological-mod
 el-gmo-clones-a-solution-to-theprecision-cosmology-dilemma/
SUMMARY:Do we need a new cosmological model? GMO-CLONES\, a solution to the
 \nprecision cosmology dilemma
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Jenny S
 orce : To unveil the nature of 95% of the Universe\, missions such as Eucl
 id aim at reaching a few percent precision. In\nthis quest for precision\,
  tensions between the standard cosmological model and observations already
  arise: local and global H0\nmeasurements are incompatible at more than 3
 Ïƒ\, anomalies emerge within the CMB\, etc. These tensions suggest that 
 we should perhaps not be so quickly inclined to disregard our observationa
 l site as a bias factor: Accuracy\nis not Precision.  Few percent precisio
 n and local-induced biases are of the same order of magnitude.  A precise\
 nmapping of the local distribution of matter is essential to properly acco
 unt for these biases. Simulations constrained to resemble the local Univer
 se constitute the tool of choice for such a mapping. I will summarize the 
 genesis of the initial conditions of such simulations as well as present a
  few results that promise to tremendously impact our understanding of the 
 local-induced biases that will matter in future analyses. Eventually\, I w
 ill present the initial conditions of the\nGMO-CLONES (GMO-Constrained LOc
 al &amp\; Nesting Environment Simulation) suite to reach an Accurate Preci
 sion Cosmology.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1473@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20190131T104500
DTEND;TZID=Europe/Paris:20190131T114500
DTSTAMP:20210605T131848Z
URL:https://iecl.univ-lorraine.fr/events/quelques-proprietes-geometriques-
 des-graphes-stables/
SUMMARY:Quelques propriétés géométriques des graphes stables
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Bénéd
 icte Haas : Considérons un graphe G_n uniformément choisi dans\nl'ensemb
 le des graphes à  n noeuds étiquetés avec des degrés D_1\,â€¦\,D_
 n\ndonnés\, eux-mêmes aléatoires i.i.d. tels que E[D^2]&lt\;âˆž et P
 (D=2)E[D]. On se place ici dans le cas critique\nE[D(D-1)]=E[D] et on supp
 ose que P(D=k)âˆ¼ck^{-2-Î±}\, 1&lt\;Î±&lt\;2. Des travaux de\nJosep
 h 14\, Riordan 12 et Conchon-Kerjan et Goldschmidt (à  paraître)\, il\n
 résulte que le graphe G_n\, après normalisation\, converge en loi vers u
 n\ngraphe continu aléatoire appelé graphe stable d&#039\;indice Î±. No
 us\nprésenterons ici quelques propriétés géométriques de ce graphe li
 mite.\n\nBasé sur un travail en collaboration avec C. Goldschmidt et D.\n
 Sénizergues.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1460@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20190207T104500
DTEND;TZID=Europe/Paris:20190207T114500
DTSTAMP:20210605T131846Z
URL:https://iecl.univ-lorraine.fr/events/multiple-fragmentation-stochastic
 -processes-driven-by-a-spatial-flow/
SUMMARY:Multiple-fragmentation stochastic processes driven by a spatial flo
 w
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Lucian 
 Beznea : We study stochastic multiple-fragmentation processes driven by a 
 spatial flow.  The final goal is actually to make a numerical simulation o
 f the time evolution of a system of particles located on an Euclidean surf
 ace. \n\nWe take into account not only the fragmentation of the mass of a 
 particle\, but also of the kinetic energy.  The talk is based on a  joint 
 work with Ioan R. Ionescu and Oana Lupascu-Stamate.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1470@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20190228T104500
DTEND;TZID=Europe/Paris:20190228T114500
DTSTAMP:20210605T131847Z
URL:https://iecl.univ-lorraine.fr/events/segregation-urbaine-distances-foc
 ales-et-effets-de-distorsion/
SUMMARY:Ségrégation urbaine: distances focales et effets de distorsion
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Madalin
 a Olteanu : Nous proposons une méthode d'analyse des dissimilarités spat
 iales d'une ville fondée sur la représentation de celle-ci par un faisce
 au de trajectoires\, obtenues en explorant la ville à  partir de chacun 
 de ses points. L'échelle à  partir de laquelle une trajectoire converge
  vers la ville entière constitue en quelque sorte une distance focale : l
 e rayon du disque qu'il faut parcourir\, en partant de tel point\, pour Â
 « voir Â» la ville telle qu'elle est en réalité\, dans son ensemble. 
 Cette distance dépend de la variable (ou de la distribution) considérée
 \, ainsi que du seuil de convergence choisi. Une intégrale permet à  la
  fois de s'affranchir de l'arbitraire dans le choix du seuil et d'identifi
 er les points pour lesquels la convergence est presque toujours lente\, y 
 compris pour des seuils relativement élevés. Nous définissons ainsi un 
 coefficient de distorsion\, qui mesure à  quel point l'image de la ville
 \, perçue en tel ou tel point\, est différente de son image globale rée
 lle. Travail en collaboration avec J. Randon-Furling (Université Paris 1)
  et W. Clark (UCLA)
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1469@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20190307T104500
DTEND;TZID=Europe/Paris:20190307T114500
DTSTAMP:20210605T131847Z
URL:https://iecl.univ-lorraine.fr/events/approximation-avec-extinction-de-
 processus-de-markov-immortels/
SUMMARY:Approximation avec extinction de processus de Markov immortels
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Edouard
  Strickler : De  nombreux modèles écologiques sont représentés par des
  équations différentielles ordinaires. Si ces modèles nous permettent\,
  plus ou moins facilement de comprendre certains comportements observés d
 ans la nature\, ils ne prennent pas en compte deux éléments inhérents 
 à  la vie réelle : l'aléa et la tragique destinée de toute population
  - la mort en temps fini.\nDans cet exposé\, nous considérons un process
 us de Markov X "immortel" et une famille de processus de Markov X^N qui me
 urent en temps fini\, et qui convergent vers X\, et nous explorerons le co
 mportement de la famille des distributions quasi-stationnaires (QSD) assoc
 iées aux X^N. Nous verrons en particulier que ce comportement dépend for
 tement de la nature du processus X (persistant ou non). Cela permet\, en u
 n certain sens\, de justifier l'approximation et l'étude de processus imm
 ortels.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1467@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20190314T104500
DTEND;TZID=Europe/Paris:20190314T114500
DTSTAMP:20210605T131847Z
URL:https://iecl.univ-lorraine.fr/events/recurrence-quantitative-de-certai
 ns-systemes-dynamiques-preservant-une-mesure-infinie-en-dimesion-un/
SUMMARY:Récurrence quantitative de certains systèmes dynamiques préserva
 nt une mesure infinie en dimesion un
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Nasab Y
 assine : Nous nous intéressons au comportement asymptotique du premier te
 mps de retour des orbites d'un système dynamique dans un petit voisinage 
 de leurs points de départ. Nous étudions cette quantité dans le context
 e de systèmes dynamiques préservant une mesure infinie. Plus préciséme
 nt\, nous considérons le cas de $mathbb{Z}$-extensions de sous-shift de t
 ype fini. Nous considérons également un modèle probabiliste pour éclai
 rer la stratégie de nos preuves.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1472@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20190404T104500
DTEND;TZID=Europe/Paris:20190404T114500
DTSTAMP:20210605T131847Z
URL:https://iecl.univ-lorraine.fr/events/estimating-finite-mixtures-of-sem
 i-markov-chains-an-application-to-the-segmentation-of-temporal-sensory-dat
 a/
SUMMARY:Estimating finite mixtures of semi-Markov chains: an application to
  the segmentation of temporal sensory data
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Hervé 
 Cardot : In food science\, it is of great interest to get information abou
 t the temporal perception of aliments  to create new products\, to modify 
  existing ones or more generally to understand the perception mechanisms. 
 Temporal Dominance of Sensations (TDS) is a technique to measure temporal 
 perception which consists in choosing sequentially attributes describing a
  food product over tasting. \nThis work introduces new statistical models 
 based on finite mixtures of semi-Markov chains  in order to describe data 
 collected with the TDS protocol\, allowing different temporal perceptions 
 for a same product within a population. The identifiability of the paramet
 ers of such mixture models is discussed.  Sojourn time distributions are f
 itted with gamma probability distribution and a penalty is added to the lo
 g likelihood to ensure convergence of the EM algorithm to a non degenerate
  solution.  Information criterions are employed for determining the number
  of mixture components. Then\, the individual qualitative trajectories are
  clustered  with the help of the maximum a posteriori probability (MAP) ap
 proach. A simulation study confirms the good behavior of the proposed esti
 mation procedure. The methodology is illustrated on an example of consumer
 s perception of a Gouda cheese and assesses the existence of several behav
 iors in terms of  perception of this product. \nJoint work with G. Lecuell
 e\, P. Schlich and M. Visalli (Centre des Sciences du Gout et de l'Aliment
 ation\, UMR Agrosup-CNRS-INRA-UB\, Dijon)
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1443@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20190425T104500
DTEND;TZID=Europe/Paris:20190425T114500
DTSTAMP:20210605T131844Z
URL:https://iecl.univ-lorraine.fr/events/representation-de-poisson-elle-a-
 quelque-chose-de-plus-que-les-autres-nont-pas/
SUMMARY:Représentation de Poisson : elle a quelque chose de plus que les a
 utres n'ont pas
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Ulysse 
 Herbach : Je vais profiter de cette occasion pour délaisser un peu la bio
 logie au profit des maths et parler d'un sujet qui me tient à  cÅ“ur 
 en ce moment : la représentation de Poisson. Introduite par C. W. Gardine
 r en 1977 comme un ansatz pratique pour résoudre certaines équations ma
 îtresses (alias Kolmogorov progressives) représentant des systèmes de r
 éactions chimiques modélisés par des processus markoviens de sauts\, ce
 tte représentation a fait ses preuves d'un point de vue formel mais n'a p
 as encore livré tous ses secrets mathématiques.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1468@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20190509T104500
DTEND;TZID=Europe/Paris:20190509T114500
DTSTAMP:20210605T131847Z
URL:https://iecl.univ-lorraine.fr/events/une-nouvelle-famille-de-couplages
 -martingale-en-dimension-un/
SUMMARY:Une nouvelle famille de couplages martingale en dimension un
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Benjami
 n Jourdain : Nous présenterons un nouveau couplage martingale entre deux 
 mesures de probabilité $mu$ et $nu$ dans l'ordre convexe en dimension un.
  Ce couplage s'exprime explicitement en fonction des intégrales des parti
 es positive et négative de la différence des fonctions quantiles de $mu$
  et $nu$. L'intégrale de $|y-x|$ contre ce couplage est plus petite que d
 eux fois la distance de Wasserstein d'indice un entre $mu$ et $nu$. Lorsqu
 e le couplage comonotone entre $mu$ et $nu$ est donné par une application
  de transport $T$\, il minimise l'intégrale de $|y-T(x)|$ parmi tous les 
 couplages martingales. Il fait partie de toute une famille de couplages ma
 rtingales qui partage ces propriétés.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1471@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20190516T104500
DTEND;TZID=Europe/Paris:20190516T114500
DTSTAMP:20210605T131847Z
URL:https://iecl.univ-lorraine.fr/events/modele-de-tessellation-aleatoire-
 pour-la-simulation-de-paysages-agricoles/
SUMMARY:Modèle de tessellation aléatoire pour la simulation de paysages a
 gricoles
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Katarzy
 na Adamczyk : La structure du paysage agricole peut avoir une influence su
 r les processus qui s'y déploient : la propagation des maladies des cultu
 res\, le flux de gènes entre des variétés cultivées\, le transfert des
  contaminants dans le sol... L'étude des interactions entre ces processus
  et le paysage a été facilité grâce aux développement récent des sim
 ulateurs de paysages\, basés sur des modèles de la géométrie stochasti
 que.\n\nLe modèle gibbsien de tessellation aléatoire en T [1] s'inscrit 
 dans ce contexte. En effet\, le choix approprié des composantes de la fon
 ction d'énergie du modèle conduit à  des simulations réalistes des pa
 rcellaires agricoles. Dans mon exposé je présenterai le modèle et l'alg
 orithme de simulation associé. Je parlerai de la méthode d'estimation de
 s paramètres et je l'appliquerai aux données de la petite région agrico
 le en Loir-et-Cher. Enfin\, j'aborderai la question du choix de modèle qu
 i fait l'objet de nos travaux récents.\nKatarzyna Adamczyk-Chauvat\, Moun
 a Kassa\, Kiên Kiêu\, Radu Stoica.\n[1] K. Kiêu\, K. Adamczyk-Chauvat\,
  H. Monod\, R. S. Stoica. A completely random T-tessellation model and Gib
 bsian extensions. Spatial Statistics\, 6\, 118-138\, 2013.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1446@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20190523T104500
DTEND;TZID=Europe/Paris:20190523T114500
DTSTAMP:20210605T131844Z
URL:https://iecl.univ-lorraine.fr/events/spatial-interaction-modeling-of-s
 tar-clusters/
SUMMARY:Spatial interaction modeling of star clusters
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Radu Da
 n Capitan : Spatial modeling in Astronomy can benefit from visualization t
 echniques and spatial manipulation using Geographic Information Systems (G
 .I.S.) techniques. Preliminary analysis of stellar population characterist
 ics in M83 galaxy using the emitted radiation source in different waveleng
 ths shows a spatial aggregation model of stellar populations and clusters 
 that are hierarchically organized in proximity of galaxy's spiral arms. A 
 review on stellar population and cluster characteristics (Krumholz et al.\
 , 2018) shows that their definition and organization characteristics vary 
 on orders of several magnitudes\, therefore the need to analyze clustering
  models using several  mathematical models as well as G.I.S. clusters tool
 s are foreseen here  (a set of weighted features\, identifies statisticall
 y significant hot spots\, cold spots\, and spatial outliers using the Anse
 lin Local Moran's I statistic). We aim to check if the clustering is relat
 ed to a theoretical model (the density wave theory)\, where local to regio
 nal organization of stellar populations that belong to main sequence can f
 orm the hierarchically organized structures along the M83 galaxy's spiral 
 arms.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1442@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20190603T094500
DTEND;TZID=Europe/Paris:20190603T184000
DTSTAMP:20210605T131844Z
URL:https://iecl.univ-lorraine.fr/events/51es-journees-de-statistique/
SUMMARY:51es JOURNEES DE STATISTIQUE
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 http://
 www.jds2019.sfds.asso.fr : Ce congrès annuel de la Société Française d
 e Statistique\, qui va durer toute la seamine\,
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1445@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20190613T104500
DTEND;TZID=Europe/Paris:20190613T114500
DTSTAMP:20210605T131844Z
URL:https://iecl.univ-lorraine.fr/events/equivalence-orbitale-stable/
SUMMARY:Équivalence orbitale stable
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Samuel 
 Petite : Le but de cet exposé est de présenter une introduction à  un 
 thème de\nthéorie ergodique : l'équivalence orbitale stable. Ce problè
 me issu de\nla classification des algèbres de von Neumann\, se propose d'
 étudier les\nsystèmes dynamiques modulo une relation plus faible que la 
 conjugaison:\nl'équivalence orbitale. Deux tels systèmes sont orbitaleme
 nt équivalent\ns'il y a un isomorphisme préservant les orbites de chacun
  des systèmes.\nNous verrons comment les principaux invariants ergodiques
  (entropie\,\nmesure invariante\, spectre\,â€¦) varient au sein d'une 
 même classe\nd'équivalence orbitale.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1441@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20190919T104500
DTEND;TZID=Europe/Paris:20190919T114500
DTSTAMP:20210605T131843Z
URL:https://iecl.univ-lorraine.fr/events/zero-temperature-limit-for-the-br
 ownian-directed-polymer-among-poissonian-disasters/
SUMMARY:Zero temperature limit for the Brownian directed polymer among Pois
 sonian disasters
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Ryoki F
 ukushima : I am going to talk about a continuum model of directed polymer 
 in random environment. The law of the polymer is defined as the Brownian m
 otion conditioned to survive among space-time Poissonian disasters. This m
 odel is well-studied in the positive temperature regime. However\, at zero
 -temperature\, even the existence of the free energy has not been proved. 
 Stefan Junk and I have proved that the free energy exists and is continuou
 s at zero-temperature.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1450@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20191003T104500
DTEND;TZID=Europe/Paris:20191003T114500
DTSTAMP:20210605T131845Z
URL:https://iecl.univ-lorraine.fr/events/le-graphe-aleatoire-split-and-dri
 ft/
SUMMARY:Le graphe aléatoire 'split-and-drift'
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Franço
 is Bienvenu : Nous étudions un graphe aléatoire motivé par des question
 s de biologie\névolutive. Ce graphe aléatoire est défini comme la distr
 ibution\nstationnaire d'une chaîne de Markov soumise aux deux transitions
 \nsuivantes : la duplication de sommet\, lors de laquelle\, à  taux cons
 tant\,\nun sommet est choisi uniformément\, déconnecté de tous ses vois
 ins puis\nreconnecté à  un autre sommet choisi uniformément ainsi qu'a
 ux voisins de\nce second sommet\; et la perte d'arêtes\, qui consiste à
   faire disparaître\nles arêtes du graphe à  taux constant par arête
 .\n\nUne approche coalescente permet d'obtenir des formules explicites pou
 r\nles premiers moments de plusieurs invariants de graphe tels que le\nnom
 bre d'arêtes ou le nombre de sous-graphes complets d'ordre k. Nous\ndonne
 rons également une formule explicite pour la distribution des\ndegrés et
  des bornes asymptotiques sur le nombre de composantes connexes.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1447@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20191010T104500
DTEND;TZID=Europe/Paris:20191010T114500
DTSTAMP:20210605T131844Z
URL:https://iecl.univ-lorraine.fr/events/trajectoires-rugueuses/
SUMMARY:Trajectoires rugueuses
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 GdR TRA
 G : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1444@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20191017T104500
DTEND;TZID=Europe/Paris:20191017T114500
DTSTAMP:20210605T131844Z
URL:https://iecl.univ-lorraine.fr/events/generalization-guarantees-via-pac
 -bayesian-framework/
SUMMARY:Generalization guarantees via PAC-Bayesian framework
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Vera Sh
 alaeva : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1440@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20191024T104500
DTEND;TZID=Europe/Paris:20191024T114500
DTSTAMP:20210605T131843Z
URL:https://iecl.univ-lorraine.fr/events/value-distribution-of-the-riemann
 -zeta-function-and-related-probabilistic-models/
SUMMARY:Value distribution of the Riemann zeta function and related probabi
 listic models
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Ashkan 
 Nikeghbali : In this talk we shall review a few well known problems relate
 d to the distribution of the values of the Riemann zeta function on the cr
 itical line and see how one can use probabilistic models and ideas to hope
 fully gain better insights  into such problems. The main focus of the talk
  will be on the GUE conjecture and the convergence of some random analytic
  functions (coming from random matrix theory).
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1456@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20191107T104500
DTEND;TZID=Europe/Paris:20191107T114500
DTSTAMP:20210605T131845Z
URL:https://iecl.univ-lorraine.fr/events/estimation-non-parametrique-pour-
 le-taux-de-saut-dun-systeme-de-neurones-en-interactions/
SUMMARY:Estimation non paramétrique pour le taux de saut d'un système de 
 neurones en interactions
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Nathali
 e Krell : Je vais vous parler d'un travail effectué en collaboration avec
  Pierre Hodara and Eva Löcherbach. On s'intéresse à  un processus de H
 awkes à  mémoire variable. On considère un modèle de neurones en inte
 raction oà¹ le potentiel des membrane des neurones est décrit comme un 
 Processus de Markov déterministe par morceaux (notés PDMP) à  valeurs 
 dans $mathbb{R}^N\, $ oà¹ $ N$ décrit le nombre de neurones. Un drift d
 éterministe attire chaque potentiel de la membrane du neurone vers un pot
 entiel à  l'équilibre $m$. Lorsqu'un neurone saute\, le potentiel de sa
  membrane est réinitialisé à  zéro et les autres gagnent $frac{1}{N}.
 $ On s'intéresse à  l'estimation du taux de sauts d'un neurone basée s
 ur l'observation du potentiel des $N$ neurones observés jusqu'à  un tem
 ps $t$. On va étudier un estimateur à  noyaux Nadaraya-Watson pour le t
 aux de sauts et on va établir la vitesse de convergence dans $L^2 .$
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1455@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20191114T104500
DTEND;TZID=Europe/Paris:20191114T114500
DTSTAMP:20210605T131845Z
URL:https://iecl.univ-lorraine.fr/events/an-exploratory-spatial-analysis-o
 f-access-to-physical-and-digital-retail-banking-channels-in-the-uk/
SUMMARY:An Exploratory Spatial Analysis of Access to Physical and Digital R
 etail Banking Channels in the UK
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Andrea 
 Sonea : Having access to one's own money is facilitated by banking channel
 s\, namely\; branches\, ATMs\, call centres\, online and mobile banking as
  well as the post office. Whilst banks provide channels\,  individual expe
 riences of access are not homogenous. Indeed\, socio-economic and geograph
 ical considerations lead to vastly varied access. With digital banking hav
 ing permeated society\, individuals across the UK use physical network poi
 nts of access - such as branches or ATMs - less than previously.  As such\
 , the cost of running such network points have become relatively more expe
 nsive and banks have started retreating their physical access point presen
 ce. Whilst this has been anecdotally reported in mainstream media\, there 
 has been no quantitative analysis on the impact of said closures. \n\nAs s
 uch\, we propose a simple methodology to identify critical points of acces
 s. Our hope is that regulators and industry players can then use this fram
 ework to ascertain integral nodes the the UK's banking channel network. Th
 e framework considers all channels available to an individual\, in respect
  to their place of residence. The distance to the closest point of access 
 is calculated as well as the impact of the closure of that access point. T
 he impact is measured as the difference in distance between the closest an
 d second-closest point of access\, reflecting an increase in the difficult
 y of banking access\n\nExploratory spatial data analysis at both UK and re
 gional level showed strong spatial patterns of the points of access to ban
 king \; significant rural/urban clusters could be identified as well as a 
 North/South divide which we need to explore further. No significant associ
 ation was found between distance metrics and income and employment. Despit
 e data limitations\, the indicators used in this study can be used to iden
 tify  areas vulnerable to the closure of the last points of access. We lea
 rned that the majority of the infrastructure for access is no longer opera
 ted by banks. In this context\, it becomes even more critical to maintain 
 and monitor a dynamic map of access and therefore we recommend more transp
 arency on location\, capability and capacity of the points of access from 
 all players\, as well as on broadband availability and quality from teleco
 m providers. Retail banking access should be treated as a joined-up system
  so that territorial coverage can be ensured\, such that entire communitie
 s are not accidentally excluded from participation in the economy.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1426@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20191121T104500
DTEND;TZID=Europe/Paris:20191121T114500
DTSTAMP:20210605T131842Z
URL:https://iecl.univ-lorraine.fr/events/une-condition-pratique-pour-lergo
 dicite-exponentielle-de-processus-de-markov-deterministes-par-morceaux/
SUMMARY:Une condition "pratique" pour l'ergodicité exponentielle de Proces
 sus de Markov Déterministes par Morceaux
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Édouar
 d Strickler : Les Processus de Markov Déterministes par Morceaux (PDMP) s
 ont des processus obtenus par la modulation aléatoire d'un nombre fini de
  champs de vecteurs. Il y a quelques années\, deux conditions de type Hö
 rmander portant sur les crochets de Lie des champs de vecteurs ont été d
 onnées pour vérifier l'unicité d'une probabilité invariante (condition
  Â« faible Â») ainsi que l'ergodicité exponentielle (condition Â« f
 orte Â»). En principe\, la condition Â« faible Â» ne suffit pas pour
  vérifier l'ergodicité exponentielle du PDMP. Néanmoins\, nous verrons 
 dans cet exposé\, après avoir rappelé ces conditions\,que s'il existe u
 n point accessible annulant une combinaison linéaire des champs de vecteu
 rs\, la condition Â« faible Â» est suffisante pour l'ergodicité expon
 entielle. Une application sera donnée à  des systèmes de Lotka-Volterr
 a switchés. Basé sur un travail avec Michel Benaïm et Tobias Hurth.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1448@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20191128T104500
DTEND;TZID=Europe/Paris:20191128T114500
DTSTAMP:20210605T131844Z
URL:https://iecl.univ-lorraine.fr/events/processus-de-naissances-et-morts-
 en-grande-populationechelles-de-temps-distribution-quasi-stationnaire-et-r
 esilience/
SUMMARY:Processus de naissances et morts en grande population\,\néchelles 
 de temps\, distribution quasi stationnaire et résilience
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Pierre 
 Collet : Avec S.Meleard et J.-R.Chazottes nous avons étudié des\nprocess
 us de naissances et morts d'une ou plusieurs espèces qui\ndépendent d'un
  (grand) paramètre donnant une échelle pour la taille de\nla population.
   En supposant qu'il existe un unique point fixe\nglobalement attractant p
 our le système dynamique (normalisé) en\npopulation infinie\, nous étab
 lissons des bornes sur le temps\nd'extinction global ainsi que l'existence
  et une borne supérieure sur\nle temps de convergence vers la mesure quas
 i stationnaire. Avec\nS.Martinez nous avons utilisé ces résultats pour 
 établir une relation\nmicro-macro qui permet de déterminer la résilienc
 e à   partir des\nfluctuations d'une trajectoire du processus.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1428@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20191212T104500
DTEND;TZID=Europe/Paris:20191212T114500
DTSTAMP:20210605T131842Z
URL:https://iecl.univ-lorraine.fr/events/comparaison-de-survie-asymptotiqu
 e-pour-les-processus-discontinusune-etape-cle-pour-la-quasi-stationarite/
SUMMARY:Comparaison de survie asymptotique pour les processus discontinus\,
 \n\nune étape clé pour la quasi-stationarité
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Auréli
 en Velleret : Cet exposé portera sur la dynamique en temps long de certai
 ns processus de Markov homogènes en temps. Pour ces processus auxquels so
 nt associés des événements d'extinction\, notre intérêt se tournera v
 ers les trajectoires survivantes. La première question que j'aborde est l
 'existence et la possible unicité d'une distribution quasi-stationaire. J
 e reprends pour cela l'approche de Champagnat et Villemonais\, dont j'esqu
 isserai le lien avec la récurrence de Harris. Une étape clé de la preuv
 e est alors de savoir comparer la survie à  long terme entre différente
 s conditions initiales.\n\nPour les processus en temps et espace continus 
 à  trajectoires discontinues\, un argument spécifique est nécessaire p
 our gérer les trajectoires pathologiques (problématiques à  court term
 e mais non représentatives à  long terme). La méthode que je vais vous
  décrire est sans doute bien technique\, mais avec sa flexibilité\, elle
  me semble très efficace. Je l'illustrerai sur quelques exemples : proces
 sus à  sauts purs ou déterministes par morceaux\, voire combinant des s
 auts à  une diffusion.\n\nSi le temps le permet\, j'évoquerai aussi un 
 lien direct entre quasi-stationarité\, quasi-ergodicité et grandes dévi
 ations.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1449@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20191219T104500
DTEND;TZID=Europe/Paris:20191219T114500
DTSTAMP:20210605T131844Z
URL:https://iecl.univ-lorraine.fr/events/on-the-convex-hull-of-several-gau
 ssian-random-walks-in-higher-dimensions/
SUMMARY:On the convex hull of several Gaussian random walks in higher dimen
 sions
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Julien 
 Randon-Furling : We derive explicit formulae for the expected volume and t
 he expected number of faces of the convex hull of several multidimensional
  Gaussian random walks\, in terms of the Gaussian persistence probabilitie
 s. We generalize further our formulae to Gaussian random walks with random
  (Gaussian) starting points. Special cases include the d-dimensional Gauss
 ian polytope with or without the origin.\n\nJoint work with Dmitry Zaporoz
 hets (Steklov Institute St Petersburg)
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1452@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20200109T104500
DTEND;TZID=Europe/Paris:20200109T114500
DTSTAMP:20210605T131845Z
URL:https://iecl.univ-lorraine.fr/events/stochastic-analysis-of-the-neutro
 n-transport-equation/
SUMMARY:Stochastic Analysis of the Neutron Transport Equation
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Emma Ho
 rton : The neutron transport equation (NTE) describes the net movement of 
 neutrons through an inhomogeneous fissile medium\, such as a nuclear react
 or. One way to derive the NTE is via the stochastic analysis of a spatial 
 branching process. This approach has been known since the 1960/70s\, howev
 er\, since then\, very little innovation in the literature has emerged thr
 ough probabilistic analysis. In recent years\, however\, the nuclear power
  and nuclear regulatory industries have a greater need for a deep understa
 nding the spectral properties of the NTE.\n \nIn this talk I will formally
  describe the dynamics of the so-called neutron branching process (NBP)\, 
 along with an associated Feynman Kac representation. I will then discuss h
 ow the latter can be used to consider the long-term behaviour of the nucle
 ar fission processes through both a Perron-Frobenius decomposition and a s
 trong law of large numbers result.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1429@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20200116T104500
DTEND;TZID=Europe/Paris:20200116T114500
DTSTAMP:20210605T131842Z
URL:https://iecl.univ-lorraine.fr/events/from-generative-models-of-protein
 -sequences-to-evolution-guided-protein-design/
SUMMARY:From generative models of protein sequences to evolution-guided pro
 tein design
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Martin 
 Weigt : Thanks to the sequencing revolution in biology\, protein sequence\
 ndatabases have been growing exponentially over the last years.\nData-driv
 en computational approaches are becoming more and more\npopular in explori
 ng this increasing data richness. In my talk\, I will\nshow that global st
 atistical modeling approaches\, like (Restricted)\nBoltzmann Machines are 
 able to accurately capture the natural\nvariability of amino-acid sequence
 s across entire families of\nevolutionarily related but distantly diverged
  proteins. We show that\nthese models are biologically interpretable\; the
 y allow to extract\ninformation about the three-dimensional protein struct
 ure and about\nprotein-protein interactions from sequence data\, and they 
 unveil\ndistributed sequence motifs. These models can be seen as highly\np
 erformant generative models - they capture the natural sequence\nvariabili
 ty far beyond fitted quantities\, and they allow to design\nnovel\, fully 
 functional proteins by simple MCMC sampling approaches.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1453@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20200130T104500
DTEND;TZID=Europe/Paris:20200130T114500
DTSTAMP:20210605T131845Z
URL:https://iecl.univ-lorraine.fr/events/universalite-dans-les-modeles-ave
 c-contraintes-cinetiques-le-role-desbarrieres-denergie/
SUMMARY:Universalité dans les modèles avec contraintes cinétiques : le r
 ôle des\nbarrières d'énergie
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Laure M
 arêché : Les modèles avec contraintes cinétiques constituent une class
 e de\nmodèles de mécanique statistique qui ont été introduits par les\
 nphysiciens pour décrire le comportement du verre. Il s'agit de modèles\
 nde configurations sur des graphes dans lesquels chaque sommet du graphe\n
 est soit à  l'état 0\, soit à  l'état 1\, et ne peut changer d'état
  que si\nune contrainte de la forme Â« il y a assez de zéros dans le vo
 isinage du\nsommet Â» est satisfaite. Il existe une infinité de contrai
 ntes\npossibles\, et les propriétés d'un modèle dépendent fortement du
  choix de\nsa contrainte. Une question très importante est donc celle de\
 nl'universalité : peut-on répartir cette infinité de modèles en un nom
 bre\nfini de classes selon leur comportement ? Cette question a récemment
  été\nrésolue lorsque le graphe de base est Z2 pour une classe de modè
 les plus\nsimple\, la percolation bootstrap\, que l'on peut considérer co
 mme une\nversion déterministe et monotone des modèles avec contraintes\n
 cinétiques. Cependant\, les modèles avec contraintes cinétiques\nprése
 ntent un phénomène de barrière d'énergie qui peut rendre leur\ncomport
 ement très différent de celui de la percolation bootstrap\, et\nnécessi
 tent donc une classification d'universalité plus fine. Dans cet\nexposé\
 , on présentera une telle classification d'universalité pour les\nmodèl
 es avec contraintes cinétiques.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1454@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20200206T104500
DTEND;TZID=Europe/Paris:20200206T114500
DTSTAMP:20210605T131845Z
URL:https://iecl.univ-lorraine.fr/events/stabilite-du-theoreme-de-bakry-em
 ery/
SUMMARY:Stabilité du théorème de Bakry-Emery
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Max Fat
 hi : Le theoreme de Bakry-Emery indique que\, sous une condition d'uniform
 e\nconvexité du potentiel\, certaines mesures de probabilités vérifient
  une\ninégalité de Poincaré\, avec une constante meilleure que celle as
 sociée à \nla mesure gaussienne. De manière équivalente\, ce résulta
 t s'interprète\ncomme une borne sur les valeurs propres de certains opér
 ateurs de\ndiffusion. Dans cet exposé\, je présenterai un résultat de s
 tabilité : si\nune telle mesure a une constante de Poincaré proche de ce
 lle de la\ngaussienne\, alors elle contient presque un facteur gaussien\, 
 avec des\nbornes d'erreur explicites. La preuve repose sur une combinaison
 \nd'arguments élémentaires de calcul des variations\, et de la méthode 
 de\nStein sur l'estimation de distances entre mesures de probabilités. Co
 mme\napplication\, on obtient des formes inverses de certaines inégalité
 s de\nconcentration pour les mesures uniformément log-concaves. Travail e
 n\ncollaboration avec Thomas Courtade.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1430@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20200213T104500
DTEND;TZID=Europe/Paris:20200213T114500
DTSTAMP:20210605T131842Z
URL:https://iecl.univ-lorraine.fr/events/nearest-neighbour-markov-point-pr
 ocesses-on-graphs-with-euclidean-edge/
SUMMARY:Nearest-neighbour Markov point processes on graphs with Euclidean e
 dge
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Marie-C
 olette van Lieshout : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1432@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20200227T104500
DTEND;TZID=Europe/Paris:20200227T114500
DTSTAMP:20210605T131842Z
URL:https://iecl.univ-lorraine.fr/events/projection-de-processus-ponctuels
 -determinantaux-et-applications-aux-methodes-monte-carlo/
SUMMARY:Projection de processus ponctuels déterminantaux et applications a
 ux méthodes Monte-Carlo
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Adrien 
 Mazoyer : Dans ces travaux effectués en collaboration avec J.-F. Coeurjol
 ly (UQAM\, Montréal) et P.-O. Amblard (Gipsa-Lab\, Grenoble)\, nous propo
 sons d'estimer une intégrale à  partir de points de quadrature produits
  par un processus ponctuel déterminantal (DPPs)\, construits à  partir 
 de noyaux de type Dirichlet. Sous l'hypothèse que l'intégrande appartien
 t à  un certain espace de Sobolev de régularité s&gt\;1/2 (condition v
 érifiée par de nombreuses fonctions non-continument différentiable)\, l
 'estimateur ainsi construit satisfait alors un théorème central limite a
 vec une variance explicite et une vitesse de convergence hyperuniforme. Gr
 âce à  la structure de ces DPPs\, il est également possible d'utiliser
  une même configuration de points et\, via la projection de ces points\, 
 estimer des intégrales de fonctions définies sur des espaces de dimensio
 n inférieure\, tout en conservant les résultats asymptotiques obtenus pr
 écédemment.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1427@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20200305T104500
DTEND;TZID=Europe/Paris:20200305T114500
DTSTAMP:20210605T131842Z
URL:https://iecl.univ-lorraine.fr/events/percolation-de-dernier-passage-ge
 neralisee-etude-sur-le-cylindre/
SUMMARY:Percolation de dernier passage généralisée : étude sur le cylin
 dre
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Jérôm
 e Casse : La percolation de dernier passage dirigée est\, classiquement\,
  un modèle de croissance dans le quart de plan discret. Pour croitre de l
 a case $(i\,j)$\, il faut que les cases $(i-1\,j)$ et $(i\,j-1)$ soient pr
 ésentes dans notre amas de croissance\, puis attendre un temps aléatoire
  $tau_{(i\,j)}$. Ce modèle est notemment intéressant pour modéliser le 
 temps d'asséchement d'un terrain.\n\nDans cet exposé\, je présente une 
 généralisation de la percolation de dernier passage dirigée dans le cas
  oà¹ le temps à  attendre $tau_{(i\,j)}$ dépend des temps d'arrivée 
 des cases $(i-1\,j)$ et $(i\,j-1)$ dans l'amas et je présente ce modèle 
 non pas comme un modèle de croissance dans le quart de plan\, mais dans u
 n cylindre de taille $L$. Dans le cylindre\, il apparait ainsi une ligne d
 e front pour notre amas.\n\nL'objet de cet exposé va être d'étudier deu
 x propriétés asymptotiques (en temps) de cette ligne de front: sa vitess
 e et sa forme. Nous verrons que\, dans des cas particuliers dits solubles 
 ou intégrables\, cette vitesse et cette forme ont une forme explicite en 
 fonction des paramètres du modèle. Puis\, j'expliquerai par quelle magie
  ces cas sont solubles\, alors que les autres ne les sont a priori pas.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1424@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20200312T104500
DTEND;TZID=Europe/Paris:20200312T114500
DTSTAMP:20210605T131841Z
URL:https://iecl.univ-lorraine.fr/events/diagramme-de-phase-pour-le-area-i
 nteraction-model/
SUMMARY:Diagramme de phase pour le area-interaction model
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Pierre 
 Houdebert : Une mesure de Gibbs est une mesure de probabilité\, sur l'esp
 ace des configurations\, qui est définie en prescrivant ses lois conditio
 nnelles. Ces lois conditionnelles admettent une densité\, par rapport au 
 processus ponctuel de Poisson homogène d'intensité z\, de la forme exp (
  - beta H )  avec H l'énergie de la configuration. Dans ce cadre une ques
 tion naturelle est de savoir\, pour chaque z et beta\, s'il existe une ou 
 plusieurs mesures ayant ces lois conditionnelles.\n\nDans un article réce
 nt en collaboration avec D. Dereudre (Lille) nous étudions le area-intera
 ction model. Pour ce modèle il est conjecturé que la non-unicité a lieu
  si et seulement si z = beta grand.\nNous répondons partiellement à  ce
 tte conjecture en prouvant l'unicité ou la non-unicité pour tous les par
 amètres z\, beta en dehors d'un compact.\n\nLes outils utilisés sont\, e
 ntre autre\, de la percolation et une représentation FK du modèle.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1439@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20200319T104500
DTEND;TZID=Europe/Paris:20200319T114500
DTSTAMP:20210605T131843Z
URL:https://iecl.univ-lorraine.fr/events/processus-empirique-base-sur-des-
 u-statistiques-a-deuxechantillons/
SUMMARY:Processus empirique basé sur des U-statistiques à  deux\néchant
 illons
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Davide 
 Giraudo : Après avoir introduit les U-statistiques à  deux échantillon
 s\,\nnous présenterons\nune version empirique de ces-dernières. Ceci per
 met de détecter un\npotentiel changement de loi\ndans un échantillon. No
 us allons donner des conditions suffisantes pour\nla convergence\ndes U-st
 atistiques à  deux échantillons dans un espace fonctionnel\napproprié 
 ainsi qu'une description du processus limite.\nIl s'agit d'un travail réa
 lisé en collaboration avec Herold Dehling\n(Ruhr-Universität Bochum) et 
 Olimjon Sharipov (National University of\nUzbekistan)
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1434@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20200326T091500
DTEND;TZID=Europe/Paris:20200326T101500
DTSTAMP:20210605T131843Z
URL:https://iecl.univ-lorraine.fr/events/estimation-et-validation-des-mode
 les-farima-faibles/
SUMMARY:Estimation et validation des modèles FARIMA faibles.
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Youssef
  Esstafa : Dans ce travail nous considérons\, le problème de l'analyse s
 tatistique des modèles FARIMA (Fractionally AutoRegressive Integrated Mov
 ing-Average) induits par un bruit blanc non corrélé mais qui peut conten
 ir des dépendances non linéaires très générales. Ces modèles sont ap
 pelés FARIMA faibles et permettent de modéliser des processus à  mémo
 ire longue présentant des dynamiques non linéaires\, de structures souve
 nt non-identifiées\, très générales. Relâcher l'hypothèse d'indépen
 dance sur le terme d'erreur\, une hypothèse habituellement imposée dans 
 la littérature\, permet aux modèles FARIMA faibles d'élargir considéra
 blement leurs champs d'application en couvrant une large classe de process
 us à  mémoire longue non linéaires.\n\nNous établissons les procédur
 es d'estimation et de validation des modèles FARIMA faibles. Nous montron
 s\, sous des hypothèses faibles de régularités sur le bruit\, que l'est
 imateur des moindres carrés des paramètres des modèles FARIMA(p\,d\,q) 
 faibles est fortement convergent et asymptotiquement normal. La matrice de
  variance asymptotique de l'estimateur des moindres carrés des modèles F
 ARIMA(p\,d\,q) faibles est de la forme "sandwich". Cette matrice peut êtr
 e très différente de la variance asymptotique obtenue dans le cas fort (
 i.e. dans le cas oà¹ le bruit est supposé iid). Nous proposons\, par de
 ux méthodes différentes\, un estimateur convergent de cette matrice. Une
  méthode alternative basée sur une approche d'auto-normalisation est ég
 alement proposée pour construire des intervalles de confiance des paramè
 tres des modèles FARIMA(p\,d\,q) faibles. Cette technique nous permet de 
 contourner le problème de l'estimation de la matrice de variance asymptot
 ique de l'estimateur des moindres carrés.\n\nNous accordons ensuite une a
 ttention particulière au problème de la validation des modèles FARIMA(p
 \,d\,q) faibles. Nous montrons que les autocorrélations résiduelles ont 
 une distribution asymptotique normale de matrice de covariance différente
  de celle obtenue dans le cadre des FARIMA forts. Cela nous permet de déd
 uire la loi asymptotique exacte des statistiques portmanteau et de propose
 r ainsi des versions modifiées des tests portmanteau standards de Box-Pie
 rce et Ljung-Box. Il est connu que la distribution asymptotique des tests 
 portmanteau est correctement approximée par un khi-deux lorsque le terme 
 d'erreur est supposé iid. Dans le cas général\, nous montrons que cette
  distribution asymptotique est celle d'une somme pondérée de khi-deux. E
 lle peut être très différente de l'approximation khi-deux usuelle du ca
 s fort. Nous adoptons la même approche d'auto-normalisation utilisée pou
 r la construction des intervalles de confiance des paramètres des modèle
 s FARIMA faibles pour tester l'adéquation des modèles FARIMA(p\,d\,q) fa
 ibles. Cette méthode a l'avantage de contourner le problème de l'estimat
 ion de la matrice de variance asymptotique du vecteur joint de l'estimateu
 r des moindres carrés et des autocovariances empiriques du bruit.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1435@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20200402T091500
DTEND;TZID=Europe/Paris:20200402T114500
DTSTAMP:20210605T131843Z
URL:https://iecl.univ-lorraine.fr/events/la-metastabilite-en-physique-stat
 istique/
SUMMARY:La métastabilité en physique statistique.
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Boris N
 ectoux : Considérons le processus de Langevin suramorti (Xt)tâ‰¥0 sol
 ution de l'équation\ndifférentielle stochastique sur R^d\n: dXt = âˆ
 ’âˆ‡f(Xt)dt + racine(h)dBt.\n\nC'est un processus prototypique utili
 sé pour modéliser l'évolution de systèmes\nstatistiques. La fonction f
  : R^d â†’ R est le potentiel du système et h &gt\; 0 sa tem-\npéra
 ture. Le processus de Langevin suramorti est métastable: il reste bloqué
  (piégé) dans des voisinages des minima locaux de f sur de longues péri
 odes de temps avant de s'en échapper. C'est une des raisons majeures qui 
 rend inaccessi-\nbles l'observation de transitions entre les états macros
 copiques du système ainsi que le calcul de quantités thermodynamiques pa
 r intégration directe des tra-\njectoires de (Xt)tâ‰¥0. De nombreux a
 lgorithmes ont été introduits ces dernières années pour accélérer l'
 échantillonnage de dynamiques métastables (e.g. les\n méthodes de Monte
 -Carlo cinétique et les accelerated dynamics algorithms introduits par A.
 F. Voter et al. à  Los Alamos). Ces algorithmes reposent sur des estimé
 es précises de l'évènement de sortie d'un état macroscopique Î© âŠ
 ‚ R\nd à  basse température (h&lt\;&lt\;1) et notamment sur le calcul
  asymptotique des taux de transition entre les états macroscopiques à  
 l&#039\;aide de la célèbre loi d&#039\;Eyring-\nKramers (1935). Dans cet
  exposé\, je présenterai des résultats récents marquant des avancées 
 sig-\nnificatives sur l&#039\;étude précise de l&#039\;évènement de so
 rtie d&#039\;un état macroscopique Î© pour le processus de Langevin sur
 amorti quand h &lt\;&lt\; 1\, ainsi que les nom-\nbreuses questions qui re
 stent ouvertes.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1451@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20200402T104500
DTEND;TZID=Europe/Paris:20200402T114500
DTSTAMP:20210605T131845Z
URL:https://iecl.univ-lorraine.fr/events/uniqueness-for-global-solutions-t
 o-the-semidiscrete-stochastic-heat-equation/
SUMMARY:Uniqueness for global solutions to the semidiscrete stochastic heat
  equation
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Tobias 
 Hurth : In spatial dimension &gt\; 2\, we consider the uniqueness problem 
 for global solutions to the stochastic heat equation\, discrete in space a
 nd continuous in time\, with a small Gaussian noise. A similar problem in 
 the continuous-space setting has been studied by Yuri Kifer. We will descr
 ibe and motivate the following result: Up to a time-dependent random norma
 lization\, the global solution is unique in the class of positive function
 s of subexponential growth and decay in space. The talk is based on a proj
 ect with Kostya Khanin and Beatriz Navarro Lameda.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1433@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20200409T104500
DTEND;TZID=Europe/Paris:20200409T114500
DTSTAMP:20210605T131842Z
URL:https://iecl.univ-lorraine.fr/events/au-dela-de-linegalite-de-poincare
 -de-second-ordre/
SUMMARY:Au-delà  de l'inégalité de Poincaré de second ordre
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Xiaochu
 an Yang : Dans cet exposé\, je présenterai des progrès récents dans l'
 approximation normale par la méthode de Stein et le calcul de Malliavin. 
 \nNous rappelons tout d'abord l'inégalité de Poincaré de second ordre\,
  une technique puissante qui donne des bornes d'erreur pour une approximat
 ion normale en termes de dérivées de Malliavin de second ordre. \nNous n
 ous concentrons ensuite sur les cas oà¹ l'inégalité de Poincaré de se
 cond ordre ne s'applique pas. Cela pourrait être la conséquence d'un man
 que de régularité ou\, dans certains cas\, de la non-tractabilité des s
 econds dérivés.\nCet exposé est basé sur plusieurs travaux conjoints a
 vec R. Lachièze-Rey\, I. Nourdin\, G. Peccati.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1425@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20200430T104500
DTEND;TZID=Europe/Paris:20200430T114500
DTSTAMP:20210605T131841Z
URL:https://iecl.univ-lorraine.fr/events/estimation-parametrique-du-terme-
 de-drift-pour-des-eds-fractionnaires/
SUMMARY:Estimation paramétrique du terme de drift pour des EDS fractionnai
 res.
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Maylis 
 Varvenne : Dans cet exposé\, nous présenterons de récents travaux effec
 tués en collaboration avec F.Panloup et S.Tindel sur l'estimation paramé
 trique du terme de drift pour une EDS fractionnaire additive sous des hypo
 thèses assurant l'ergodicité de l'EDS. La méthode d'estimation est en e
 ffet basée sur l'identification de la mesure invariante (à  définir da
 ns ce cadre a priori non-markovien) pour laquelle nous construisons une ap
 proximation à  partir d'observations discrètes de l'EDS. Nous donnerons
  des résultats de consistance ainsi qu'une borne non asymptotique sur l'e
 rreur quadratique moyenne.\nPour obtenir ce dernier résultat\, nous déta
 illerons des résultats d'inégalités de concentration pour les EDS fract
 ionnaires que nous avons développés dans de récents travaux.\nEnfin\, n
 ous discuterons de l'hypothèse d'identifiabilité intrinsèquement liée 
 à  la mesure invariante et nous donnerons quelques illustrations numéri
 ques.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1431@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20200604T104500
DTEND;TZID=Europe/Paris:20200604T114500
DTSTAMP:20210605T131842Z
URL:https://iecl.univ-lorraine.fr/events/skeletal-sdes-for-csbps/
SUMMARY:Skeletal SDEs for CSBPs
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Andreas
  Kyprianou : We look at at a coupled system of stochastic differential equ
 ations that describe an infinite parametric family of genealogical skeleta
 l decompositions of a general continuous state branching process (CSBP)\, 
 supercritical\, critical and subcritical. This puts into a common framewor
 k a number of known and new path decompositions of CSBPs\, including those
  which involve continuum random trees\, and allow us to connect the notion
  of Evans-O'Connell immortal particle decomposition to that of the skeleta
 l decomposition. This is joint work with Dorka Fekete (Exeter) and Joaquin
  Fontbona (U. de Chile).
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1436@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20200611T104500
DTEND;TZID=Europe/Paris:20200611T114500
DTSTAMP:20210605T131843Z
URL:https://iecl.univ-lorraine.fr/events/approches-bayesiennes-pour-les-pr
 otocoles-des-modeles-robustes-et-discriminatoires/
SUMMARY:Approches Bayesiennes pour les protocoles des modèles robustes et 
 discriminatoires
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Vincent
  Agboto : 
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1438@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20200618T104500
DTEND;TZID=Europe/Paris:20200618T114500
DTSTAMP:20210605T131843Z
URL:https://iecl.univ-lorraine.fr/events/marche-aleatoire-sur-les-complexe
 s-simpliciaux/
SUMMARY:Marche aléatoire sur les complexes simpliciaux.
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Laurent
  Decreusefond : Les complexes simpliciaux sont les généralisations des g
 raphes géométriques à  des relations non plus binaires mais aussi tern
 aires ou plus. Ce sont des objets très utilisés en analyse de données t
 opologiques. Nous construisons sur ces objets une nouvelle marche aléatoi
 re qui généralise la marche aléatoire canonique sur un graphe. Nous mon
 trons que son générateur est intimemement au Laplacian du complexe simpl
 icial\, qui est une généralisation du Laplacien de graphe. Nous nous int
 éressons ensuite au processus limite quand la densité du nombre de point
 s tend vers l'infini. Nous montrons comment utiliser cette marche pour loc
 aliser les trous de couverture dans un réseau radio.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1409@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20200625T104500
DTEND;TZID=Europe/Paris:20200625T114500
DTSTAMP:20210605T131839Z
URL:https://iecl.univ-lorraine.fr/events/chaines-de-markov-a-memoire-varia
 ble-et-marches-aleatoirespersistantes/
SUMMARY:Chaînes de Markov à  mémoire variable et marches aléatoires\np
 ersistantes
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Peggy C
 énac : Cet exposé présentera une petite zoologie de chaînes de Markov 
 à \nmémoire variable\, avec des conditions d'existence et unicité de m
 esure\ninvariante. Il sera ensuite question de marches aléatoires\npersis
 tantes\, construites à  partir de chaînes de Markov à  mémoire non\n
 bornée\, oà¹ les longueurs de sauts de la marche ne sont pas forcément
 \nintégrables. Un critère de récurrence/transience s'exprimant en\nfonc
 tion des paramètres du modèle sera énoncé. Suivront plusieurs\nexemple
 s illustrant le caractère instable du type de la marche\nlorsqu'on pertur
 be légèrement les paramètres. Les travaux décrits dans\ncet exposé so
 nt le fruit de plusieurs collaboration avec B. Chauvin\, F.\nPaccaut et N.
  Pouyanne ou B. de Loynes\, A. Le Ny et Y. Offret et A.\nRousselle.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1437@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20201001T104500
DTEND;TZID=Europe/Paris:20201001T114500
DTSTAMP:20210605T131843Z
URL:https://iecl.univ-lorraine.fr/events/processus-dornstein-uhlenbeck-ave
 c-seuil-estimation-des-parametres/
SUMMARY:Processus d'Ornstein-Uhlenbeck avec seuil : estimation des paramèt
 res
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Sara Ma
 zzonetto : Un processus d'Ornstein-Uhlenbeck avec seuil est un processus d
 'autoregression à  temps continu. Il suit une dynamique d'Ornstein-Uhnle
 nbeck au dessus ou dessous d'un seuil fixé\, pourtant à  ce seuil les c
 oefficients peuvent être discontinus. Nous considérons l'estimation par 
 (quasi)-maximum de vraisemblance des paramètres de dérive\, à  partir 
 d'observations à  temps continu ou discret. Dans le cas ergodique\, nous
  montrons consistance et vitesse de convergence en temps long et haute fr
 équence pour ces estimateurs. En se basant sur ces résultats\, nous dév
 eloppons un test heuristique pour la présence d'un seuil dans la dynamiqu
 e et nous concluons avec une application à  short term US interest rates
 . \nCeci est un travail avec Paolo Pigato.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1411@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20201008T104500
DTEND;TZID=Europe/Paris:20201008T114500
DTSTAMP:20210605T131839Z
URL:https://iecl.univ-lorraine.fr/events/population-structuree-en-age-en-e
 nvironnement-variable/
SUMMARY:Population structurée en âge en environnement variable
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Bertran
 d Cloez : Dans cet exposé\, nous intéressons à  des modèles simples d
 e croissance de population. Chaque individu possède une durée de vie al
 éatoire\, indépendante des durées de vie des autres individus\, et dont
  la loi dépend uniquement de la date de naissance de l'individu. A sa mor
 t ou durant son vivant mais de manière Poissonienne\, chaque individu don
 ne naissance à  de nouveaux individus. Pour ce modèle\, nous étudieron
 s ces processus avec deux outils différents : le processus de contour de 
 l'arbre et la théorie des semi-groupes. La première approche permettra d
 'avoir la loi du nombre d'individus\, divers résultats de conditionnement
  (comportement quasi-stationnaire\, loi de l'arbre conditionné à  l'ext
 inction ou la non-extinction etc.) ou des limites d'échelles. La deuxièm
 e approche permet de montrer une croissance exponentielle pour la taille d
 e la population ainsi que la convergence du profil des âges. Nous finiron
 s l'exposé avec quelques perspectives en statistiques.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1413@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20201015T104500
DTEND;TZID=Europe/Paris:20201015T114500
DTSTAMP:20210605T131840Z
URL:https://iecl.univ-lorraine.fr/events/metastabilite-pour-un-systeme-de-
 neurones-en-interaction/
SUMMARY:Métastabilité pour un système de neurones en interaction
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Pierre 
 Monmarché : On considère un système de N neurones\, dont le potentiel d
 e membrane évolue selon une dynamique de type interaction champ moyen. Pl
 us précisément\, pour chaque neurone\, ce potentiel décroît à  taux 
 constant\, et d'autre part est mis à  zéro lorsque le neurone se décha
 rge (émet un spike)\, ce qui entraîne également une augmentation du pot
 entiel de tous les autres neurones. Les spike surviennent à  des temps a
 léatoires\, à  un taux lamba(u) qui dépend du potentiel de membrane u.
  Quand lambda(u) est nul en 0 et dérivable alors\, quelque soit N\, le sy
 stème s'arrête presque sà»rement en temps fini\, c'est-à -dire qu'il
  n'y aura qu'un nombre fini de spike\, suivi d'une décroissance détermin
 iste du système vers 0. On verra que\, sous certaine condition\, le syst
 ème est néanmoins métastable\, au sens oà¹ les points suivants sont s
 atisfaits : 1) le système non-linéaire limite (N-&gt\;infini) converge v
 ers un unique équilibre non nul \; 2) le temps d'extinction d'un système
  fini de N neurones est exponentiellement grand en fonction de N \; 3) le 
 potentiel moyen du système s'approche rapidement d'une valeur positive co
 nstante\, et les temps de sortie de voisinages de cette valeur convergent 
 (quand N-&gt\;infini) vers la loi exponentielle (caractère sans mémoire\
 , imprévisible de ces déviations du comportement limite). Les démonstra
 tions reposent sur des méthodes de couplage.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1415@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20201105T104500
DTEND;TZID=Europe/Paris:20201105T114500
DTSTAMP:20210605T131840Z
URL:https://iecl.univ-lorraine.fr/events/marches-aleatoires-sur-des-graphe
 s-aleatoires-a-deux-communautes/
SUMMARY:Marches aléatoires sur des graphes aléatoires à  deux communaut
 és
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Anna Be
 n-Hamou : Le temps de mélange d'une marche aléatoire sur un graphe est 
 étroitement lié à  l'existence de Â«Â goulots d'étranglementÂ 
 Â» dans le graphe : intuitivement\, plus il est difficile pour la marche
  de s'échapper de certains sous-ensembles\, plus la marche met du temps 
 à  mélanger. Plusieurs résultats montrent que\, sur des graphes aléat
 oires qui sont presque sà»rement des expanseurs et donc n'ont typiquemen
 t pas de goulots d'étranglement (par exemple\, sur des graphes réguliers
  uniformes)\, la marche aléatoire mélange non seulement vite mais de fa
 çon très abrupte (on dit qu'elle présente le phénomène de cutoff). Da
 ns cet exposé\, nous verrons que l'on peut aussi obtenir des résultats s
 imilaires sur des graphes qui ne sont typiquement pas des expanseurs. Nous
  considérerons des graphes aléatoires munis d'une structure à  deux co
 mmunautés et montrerons qu'il existe un seuil pour la fraction d'arêtes 
 inter-blocs autour duquel la marche bascule d'un régime de mélange rapid
 e avec cutoff à  un régime de mélange lent sans cutoff.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1412@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20201112T104500
DTEND;TZID=Europe/Paris:20201112T114500
DTSTAMP:20210605T131840Z
URL:https://iecl.univ-lorraine.fr/events/graphes-de-dependance-ponderes-et
 -normalite-asymptotique/
SUMMARY:Graphes de dépendance (pondérés) et normalité asymptotique
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Valenti
 n Féray : Le but de cet exposé est de présenter la théorie des graphes
  de dépendance et une extension pondérée que j'ai récemment introduite
 . Ces théories donne des critères de normalité asymptotique pour des so
 mmes de variables aléatoires faiblement dépendantes. Elle s'applique en 
 particulier aux nombres de sous-structures de taille fixée dans des objet
 s combinatoires aléatoires ou des modèles de physique statistique. L'exp
 osé sera illustré par des exemples de compte de sous-mots dans des mots 
 aléatoires et des applications à  un système de particule (SSEP\, symm
 etric simple exclusion process) et au modèle d'Ising.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1416@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20201119T104500
DTEND;TZID=Europe/Paris:20201119T114500
DTSTAMP:20210605T131840Z
URL:https://iecl.univ-lorraine.fr/events/schemas-preservant-lasymptotique-
 pour-quelques-modeles-stochastiques/
SUMMARY:Schémas préservant l'asymptotique pour quelques modèles stochast
 iques
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Charles
 -Edouard Bréhier : On considère des systèmes multiéchelles d'Equations
  Différentielles Stochastiques: quand un paramètre epsilon tend vers 0\,
  la composante lente converge soit vers la solution d'une équation diffé
 rentielle ordinaire (principe de moyennisation)\, soit vers la solution d'
 une équation différentielle stochastique (approximation diffusion). L'ob
 jectif d'un schéma préservant l'aymptotique est d'être consistent pour 
 tout epsilon\, et d'admettre un schéma limite quand epsilon tend vers 0\,
  qui soit consistent avec l'équation limite au niveau continu.\n\nOn verr
 a que pour les modèles EDS la consistance du schéma limite avec l'équat
 ion limite n'est pas évidente: par exemple\, le schéma limite peut être
  naturellement associé à  une interprétation Itô du bruit\, alors que
  l'équation limite est associée à  l'interprétation Stratonovich. On 
 décrira des exemples et contre-exemples de schémas préservant l'asympto
 tique.\n\nEnfin\, on montrera (dans le régime moyennisation) une estimati
 on d'erreur uniforme par rapport à  epsilon\, en fonction du pas de temp
 s: le schéma est uniformément précis.\n\n\nIl s'agit d'un travail en co
 llaboration avec Shmuel Rakotonirina-Ricquebourg (Lyon 1).
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1414@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20201126T104500
DTEND;TZID=Europe/Paris:20201126T114500
DTSTAMP:20210605T131840Z
URL:https://iecl.univ-lorraine.fr/events/simulation-exacte-du-temps-necess
 aire-a-une-diffusion-pour-sortir-dun-intervalle/
SUMMARY:Simulation exacte du temps nécessaire à  une diffusion pour sort
 ir d'un intervalle
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Samuel 
 Herrmann : Les diffusions (famille de solutions d'équations différentiel
 les stochastiques) jouent un rôle primordial en modélisation stochastiqu
 e avec de nombreux champs d'application. Il est donc essentiel de pouvoir 
 simuler précisément les trajectoires de ces processus et toute variable 
 aléatoire qui y serait liée. Dans cette communication\, nous nous intér
 esserons en particulier au premier instant de sortie d'un intervalle donn
 é. Nous considérons donc (Xt) la solution de dXt = Î¼(Xt)dt + dBt\, X0
  âˆˆ ]a\,b[\, tâ‰¥0\, oà¹ (Bt) est un mouvement brownien et l'obj
 ectif se résume à  la simulation numérique de Ï„_ab = inf{tâ‰¥0
 : Xt âˆ‰ ]a\,b[}. Cette variable aléatoire dépend de la trajectoire 
 du processus et non simplement d'une marginale à  un temps fixé\, ce qu
 i rend plus compliquée sa simulation numérique. Une première approche c
 onsiste à  introduire des schémas numériques basés sur la discrétisa
 tion temporelle. Ces schemas permettent d'obtenir un squelette de la diffu
 sion et d'en déduire une approximation du temps de sortie. Une autre faç
 on d'appréhender le problème de simulation est d'utiliser une méthode d
 e rejet pour simuler directement et de façon exacte le temps de sortie. C
 'est cette méthode que je souhaite vous présenter. Une première étude 
 sur la simulation exacte notamment des marginales de diffusion fut introdu
 ite par Beskos et Roberts puis complétée par différents travaux par la 
 suite. En ce qui concerne les temps d'arrêt\, Cristina Zucca et moi-même
  avons étudié dans un premier temps les premiers instants de passage des
  diffusions avant de nous intéresser aux temps de sortie dont la complexi
 té (au niveau des algorithmes) est bien supérieure. Références : https
 ://iecl.univ-lorraine.fr/ProbaStat/covid/downloads/2020-11-26_Samuel_Herrm
 ann_abstract.pdf
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1423@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20201203T104500
DTEND;TZID=Europe/Paris:20201203T114500
DTSTAMP:20210605T131841Z
URL:https://iecl.univ-lorraine.fr/events/transport-optimal-martingale-et-c
 onstruction-de-couplages/
SUMMARY:Transport optimal martingale et construction de couplages
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Nicolas
  Juillet : Le problème du transport optimal de Monge\, sous sa forme Â«
 Â KantorovichÂ Â»\, se formule particulièrement bien en termes prob
 abilistes puisqu'il consiste à  minimiser l'espérance de la distance (o
 u d'une autre fonction) de deux variables aléatoires dont les marges\, le
 s fameux Â«Â déblaisÂ Â» et Â«Â remblaisÂ Â»\, sont des d
 onnées du problème. En somme il s'agit de trouver un couplage (un transp
 ort\, une loi jointe) optimal(e). Je parlerai de certains de mes travaux s
 ur la variante Â«Â martingaleÂ Â» du problème et des couplages sp
 écifiques (dernièrement d'une infinité indénombrable de lois) qui en o
 nt émergé. Des liens avec le problème de plongement de Skorokhod et cer
 taines représentations de Choquet seront évoqués. Travaux en collaborat
 ion avec Mathias Beiglböck\, et plus récemment avec Martin Huesmann et M
 artin Brà¼ckerhoff.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1417@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20201210T104500
DTEND;TZID=Europe/Paris:20201210T114500
DTSTAMP:20210605T131840Z
URL:https://iecl.univ-lorraine.fr/events/quelques-modeles-de-regression-ex
 treme/
SUMMARY:Quelques modèles de régression extrême
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Antoine
  Usseglio-Carleve : Si estimer la médiane (quantile de niveau 0.5) ou le 
 quartile (quantile de niveau 0.25 ou 0.75) d'une variable aléatoire Y par
 aît évident lorsque l'on dispose d'un échantillon de taille n\, qu'en e
 st-il si le niveau de quantile que l'on cherche à  estimer dépasse 1-1/
 n ? Dans ce cas\, l'usage de la classique statistique d'ordre renvoie syst
 ématiquement le maximum de l'échantillon\, et mène alors à  une estim
 ation non-consistante du quantile désiré. Grâce à  la théorie des va
 leurs extrêmes\, on trouve dans la littérature des méthodes d'extrapola
 tion pour estimer de tels quantiles. La particularité de ce travail est q
 ue la variable d'intérêt Y est impactée par un vecteur de covariables X
 . L'enjeu est alors d'estimer des quantiles extrêmes de la loi conditionn
 elle de Y sachant X=x. Pour cela\, on propose d'abord une approche de rég
 ression purement non-paramétrique\, en proposant des estimateurs de quant
 ile et d'expectile (une alternative au quantile que l'on introduira) extr
 êmes\, et en étudiant leurs propriétés asymptotiques. La vitesse de co
 nvergence de ces estimateurs se dégradant assez fortement lorsque la tail
 le de la covariable X augmente\, on proposera alors quelques modèles sur 
 X et Y permettant de contourner le fléau de la dimension. Quelques applic
 ations en assurance ou catastrophe naturelle seront proposées.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1410@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20201217T104500
DTEND;TZID=Europe/Paris:20201217T114500
DTSTAMP:20210605T131839Z
URL:https://iecl.univ-lorraine.fr/events/quelques-modeles-de-regression-ex
 treme-2eme-tour-apres-gros-probleme-technique/
SUMMARY:Quelques modèles de régression extrême (2ème tour après gros p
 roblème technique)
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Antoine
  Usseglio-Carleve : Si estimer la médiane (quantile de niveau 0.5) ou le 
 quartile (quantile de niveau 0.25 ou 0.75) d'une variable aléatoire Y par
 aît évident lorsque l'on dispose d'un échantillon de taille n\, qu'en e
 st-il si le niveau de quantile que l'on cherche à  estimer dépasse 1-1/
 n ? Dans ce cas\, l'usage de la classique statistique d'ordre renvoie syst
 ématiquement le maximum de l'échantillon\, et mène alors à  une estim
 ation non-consistante du quantile désiré. Grâce à  la théorie des va
 leurs extrêmes\, on trouve dans la littérature des méthodes d'extrapola
 tion pour estimer de tels quantiles. La particularité de ce travail est q
 ue la variable d'intérêt Y est impactée par un vecteur de covariables X
 . L'enjeu est alors d'estimer des quantiles extrêmes de la loi conditionn
 elle de Y sachant X=x. Pour cela\, on propose d'abord une approche de rég
 ression purement non-paramétrique\, en proposant des estimateurs de quant
 ile et d'expectile (une alternative au quantile que l'on introduira) extr
 êmes\, et en étudiant leurs propriétés asymptotiques. La vitesse de co
 nvergence de ces estimateurs se dégradant assez fortement lorsque la tail
 le de la covariable X augmente\, on proposera alors quelques modèles sur 
 X et Y permettant de contourner le fléau de la dimension. Quelques applic
 ations en assurance ou catastrophe naturelle seront proposées.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1422@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20210107T104500
DTEND;TZID=Europe/Paris:20210107T114500
DTSTAMP:20210605T131841Z
URL:https://iecl.univ-lorraine.fr/events/asymetrie-dans-la-division-cellul
 aire-etude-theorique-et-numerique/
SUMMARY:Asymétrie dans la division cellulaire\, étude théorique et numé
 rique
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Benoît
 e de Saporta : Ce travail est motivé par l'existence d'asymétrie lors de
  la division cellulaire. Après avoir examiné cette asymétrie sur des do
 nnées expérimentales\, nous introduisons un modèle probabiliste décriv
 ant les divisions successives de cellules et prenant en compte deux types 
 d'asymétrie: une asymétrie physiologique décrivant le fait que deux cel
 lules soeurs peuvent grandir à  des vitesses différentes\, et une asym
 étrie morphologique décrivant le fait que les tailles des deux cellules 
 soeurs à  la division sont différentes. Dans un premier temps\, nous ex
 pliciterons le caractère Malthusien de la dynamique\, au sens oà¹ la ta
 ille de la population croit exponentiellement tandis que la distribution d
 es tailles converge vers une distribution stable. Dans un second temps\, n
 ous étudierons les fluctuation du paramètre Malthusien en fonction des d
 ifférents paramètres du modèle. Nous montrerons que sous certaines hypo
 thèses\, l'asymétrie est optimale au sens Darwinien. Ce travail est touj
 ours en cours et est en collaboration avec Bertrand Cloez (INRAE Montpelli
 er) et Tristan Roget (Univ. Montpellier).
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1419@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20210114T104500
DTEND;TZID=Europe/Paris:20210114T114500
DTSTAMP:20210605T131841Z
URL:https://iecl.univ-lorraine.fr/events/global-sensitivity-analysis-for-m
 odels-described-by-stochastic-differential-equations/
SUMMARY:Global sensitivity analysis for models described by stochastic diff
 erential equations
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Pierre 
 Étoré : Many mathematical models involve input parameters\, which are no
 t precisely known. Global sensitivity analysis aims to identify the parame
 ters whose uncertainty has the largest impact on the variability of a quan
 tity of interest. One of the statistical tools used to quantify the influe
 nce of each input variable on the quantity of interest are the Sobol' sens
 itivity indices. In this paper\, we consider stochastic models described b
 y stochastic differential equations (SDE). We focus the study on mean quan
 tities\, defined as the expectation with respect to the Wiener measure of 
 a quantity of interest related to the solution of the SDE itself. Our appr
 oach is based on a Feynman-Kac representation of the quantity of interest\
 , from which we get a parametrized partial differential equation (PDE) rep
 resentation of our initial problem. We then handle the uncertainty on the 
 parametrized PDE using polynomial chaos expansion and a stochastic Galerki
 n projection.\n\nTalk will be in French
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1421@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20210121T104500
DTEND;TZID=Europe/Paris:20210121T114500
DTSTAMP:20210605T131841Z
URL:https://iecl.univ-lorraine.fr/events/quelques-developpements-recents-e
 n-matiere-de-gestion-et-de-surveillance-des-ressources-naturelles/
SUMMARY:Quelques développements récents en matière de gestion et de surv
 eillance des ressources naturelles
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Thibaut
  Mastrolia : Dans cet exposé\, nous étudions l'impact d'une politique de
  surveillance proposée à  un agent exploitant une ressource naturelle r
 enouvelable. Nous adoptons un modèle principal/agent en temps continu dan
 s lequel le principal conçoit un contrat\, c'est-à -dire une politique 
 de taxes/compensations\, conduisant l'agent à  un niveau d'exploitation 
 donné. Pour un contrat donné\, nous décrivons d'abord l'effort optimal 
 de l'agent en utilisant la théorie des EDSR. Sous des hypothèses de rég
 ularité sur les coefficients\, nous exprimons ensuite le contrat optimal 
 comme la solution d'une équation d'Hamilton Jacobi Bellman. Nous étendon
 s ensuite le résultat à  des coefficients non réguliers en fournissant
  des stratégies epsilon optimales à  l'aide d'un résultat d'approximat
 ion pour la fonction de valeur du régulateur. Travaux conjoints avec Idri
 s Kharroubi (Sorbonne Université) et Thomas Lim (ENSIIE).
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:1418@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20210128T104500
DTEND;TZID=Europe/Paris:20210128T114500
DTSTAMP:20210605T131840Z
URL:https://iecl.univ-lorraine.fr/events/distributions-de-tracy-widom-dord
 re-superieur/
SUMMARY:Distributions de Tracy-Widom d'ordre supérieur
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Mattia 
 Cafasso : Dans un article publié en 2018\, Le Doussal Majumdar et Schehr 
 ont introduit une famille de distributions\, indexées par un entier posit
 if n\, qui généralisent la célèbre distribution de Tracy-Widom (GUE) d
 écrivant la loi limite de la plus grande valeur propre d'une matrice alé
 atoire. Plus récemment\, les mêmes distributions sont apparues aussi dan
 s la théorie des partitions aléatoires. Après une bref introduction con
 cernant leur applications\, j'illustrerai les résultats que j'ai obtenu e
 n collaboration avec Tom Claeys et Manuela Girotti sur les grandes déviat
 ions associées à  ces distributions\, et leur liens avec les équations
  de Painlevé.
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:15@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20210211T104500
DTEND;TZID=Europe/Paris:20210211T114500
DTSTAMP:20210209T212047Z
URL:https://iecl.univ-lorraine.fr/events/estimation-non-parametrique-pour-
 des-flux-de-donnees/
SUMMARY:Estimation non paramétrique pour des flux de données
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Amir Ab
 oubacar (Université de Lille) : Dans cet exposé\, nous nous intéressero
 ns à l'estimation fonctionnelle dans un cadre non paramétrique pour des 
 flux de données. Nous donnerons une définition et une modélisation stat
 istique de ce type de données. Nous présenterons brièvement quelques qu
 estions relatives à l'estimation non paramétrique\, lorsque l'échantill
 on d'apprentissage est de nature temporelle\, spatiale ou spatio-temporell
 e et se présente sous forme de flux de données. Nous considérerons le c
 as d'un modèle statistique dans lequel la variable aléatoire générique
  est multivariée\, circulaire ou de nature fonctionnelle. Des modèles cl
 assiques seront revisités dans le contexte de flux de données\, et leurs
  propriétés asymptotiques étudiées\, notamment lorsque le processus g
 énérateur des données est stationnaire ou localement stationnaire.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de probabilités et statistique virtuelle\, IECL\, Nancy\, F
 rance
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de probabilités et statistique virtuelle:geo:
 0,0
END:VEVENT
BEGIN:VEVENT
UID:16@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20210218T104500
DTEND;TZID=Europe/Paris:20210218T114500
DTSTAMP:20210209T212928Z
URL:https://iecl.univ-lorraine.fr/events/multiple-partition-clustering/
SUMMARY:Multiple Partition Clustering
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Vincent
  Vandewalle (Université de Lille) : This talk deals with clustering when 
 several latent class variables are considered (multiple partition clusteri
 ng). Indeed\, assuming that all heterogeneity in the data can be explained
  by one single variable is very strong\, and it can be useful to consider 
 that several blocks (or linear combinations) of variables can provide diff
 erent partitions of individuals. This can reveal new lines of analysis in 
 the data. In this framework\, we present two approaches. The first one ass
 umes the existence of several groups of variables\, each leading to a diff
 erent partition of the individuals [1]. It makes it possible to classify t
 he variables into blocks\, each producing a specific grouping of individua
 ls. The model assumes the independence between blocks of variables\, and i
 n each block the independence of the variables given the cluster. An effic
 ient approach is proposed to search for the blocks of variables as well as
  performing the estimation of the different partitions of the individuals.
  The second one assumes the existence of several classifying projections i
 n the data [2]. It makes it possible to obtain different classifying proje
 ctions and the associated partitions. The model assumes that the data are 
 obtained based on linear combinations of classifying and non classifying v
 ariables\, where each classifying variable is assumed to follow a specific
  mixture distribution. The parameters of the models are estimated through 
 a generalized EM algorithm. The behavior of these models will be illustrat
 ed in simulated and real data. We will discuss how using such kind of mode
 ls can give new insight from the data analysis point of view\, and can be 
 considered for further investigation. References: [1] Marbac\, M. and Vand
 ewalle\, V. (2019). “A tractable multi-partitions clustering”. In: Com
 putational Statistics & Data Analysis 132\, pp. 167–179. [2] Vandewalle\
 , V. (2020). “Multi-Partitions Subspace Clustering”. In: Mathematics 8
 .4\, p. 597.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de probabilités et statistique virtuelle\, IECL\, Nancy\, F
 rance
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de probabilités et statistique virtuelle:geo:
 0,0
END:VEVENT
BEGIN:VEVENT
UID:46@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20210311T091500
DTEND;TZID=Europe/Paris:20210311T101500
DTSTAMP:20210302T143406Z
URL:https://iecl.univ-lorraine.fr/events/automates-cellulaires-preservant-
 un-sous-shift/
SUMMARY:Automates cellulaires préservant un sous-shift
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Samuel 
 Petite (LAMFA\, Amiens) : Les automates cellulaires forment une classe ric
 he de systèmes dynamiques sur l’ensemble des suites symboliques. Ils se
 rvent notamment de modèles simplifiés en informatique\, pour le calcul p
 arallèle\, et en physique statistique\, pour étudier l’évolution de s
 ystèmes de particules. Un problème classique consiste alors à étudier 
 les environnements laissés stable par l’évolution d’un ou plusieurs 
 automates et en particuliers leurs mesures invariantes ou les distribution
 s asymptotiques des itérés des automates sur une configuration aléatoir
 e. Nous présenterons dans cet exposé plusieurs restrictions sur ces auto
 mates en fonction de la complexité de l’environnement.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de probabilités et statistique virtuelle\, IECL\, Nancy\, F
 rance
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de probabilités et statistique virtuelle:geo:
 0,0
END:VEVENT
BEGIN:VEVENT
UID:47@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20210311T104500
DTEND;TZID=Europe/Paris:20210311T114500
DTSTAMP:20210304T171203Z
URL:https://iecl.univ-lorraine.fr/events/sous-diffusion-de-lenergie-dans-d
 es-systemes-hamiltoniens-uni-dimensionnels/
SUMMARY:Sous-diffusion de l’énergie dans des systèmes Hamiltoniens uni-
 dimensionnels
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Franço
 is Huveneers (CEREMADE\, Paris-Dauphine) : Dans cet exposé\, on s’inté
 ressera à un problème de physique statistique hors équilibre : la propa
 gation de l’énergie dans des chaînes d’oscillateurs\, classiques ou 
 quantiques\, en dimension 1. Si l’énergie est l’unique quantité cons
 ervée\, on s’attend dans la plupart des cas à observer un transport di
 ffusif. Néanmoins\, si le milieu est désordonné\, il est possible d’o
 bserver une absence totale de transport (localisation d’Anderson et loca
 lisation à N corps)\, ou un transport plus lent que diffusif\, dû à la 
 présence de goulots. J’expliquerai la phénoménologie et je donnerai u
 n modèle Hamiltonien où on peut obtenir un résultat mathématique rigou
 reux. L’exposé se base sur un travail en collaboration avec Wojciech De
  Roeck et Stefano Olla.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de probabilités et statistique virtuelle\, IECL\, Nancy\, F
 rance
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de probabilités et statistique virtuelle:geo:
 0,0
END:VEVENT
BEGIN:VEVENT
UID:45@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20210318T091500
DTEND;TZID=Europe/Paris:20210318T101500
DTSTAMP:20210301T143830Z
URL:https://iecl.univ-lorraine.fr/events/problemes-de-ruine-equation-de-la
 -chaleur-sur-un-triangle-solutions-extremales-et-jeux-a-champs-moyen/
SUMMARY:Problèmes de ruine\, équation de la chaleur sur un triangle\, sol
 utions extrémales et jeux à champs moyen
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Nabil K
 azi-Tani (ISFA\, Université Lyon 1) : Je donnerai dans cet exposé deux e
 xemples de problèmes de contrôle stochastique consistant à optimiser un
  critère discontinu\, dans lesquels d’une part\, la fonction valeur peu
 t être obtenue explicitement et d’autre part\, le contrôle optimal est
  extrémal (contrôle bang-bang). Je considèrerai d’abord le problème 
 consistant à minimiser une probabilité de ruine en temps fini pour des m
 artingales browniennes. En calculant explicitement les probabilités de so
 rties d’un triangle rectangle par le mouvement brownien (en utilisant de
 s résultats connus sur les processus de Bessel)\, il est possible de mont
 rer que la fonction valeur du problème de contrôle est une solution rég
 ulière d'une EDP de la chaleur avec des conditions aux bords discontinues
 . J’expliquerai en quoi ce problème est utile en assurance\, en biologi
 e\, ou encore en science politique. Dans un 2e temps\, je montrerai commen
 t obtenir des résultats similaires dans des problèmes de jeux différent
 iels à N joueurs\, dont je prendrai une approximation de type champs moye
 n dans le régime où N est grand. Cet exposé s’appuie sur des travaux 
 en collaboration avec Stefan Ankirchner (Jena)\, Christophette Blanchet-Sc
 alliet (Lyon)\, Julian Wendt (Jena) et Chao Zhou (Hong Kong).
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de probabilités et statistique virtuelle\, IECL\, Nancy\, F
 rance
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de probabilités et statistique virtuelle:geo:
 0,0
END:VEVENT
BEGIN:VEVENT
UID:18@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20210318T104500
DTEND;TZID=Europe/Paris:20210318T114500
DTSTAMP:20210309T085937Z
URL:https://iecl.univ-lorraine.fr/events/nouveaux-developpements-en-statis
 tique-grace-a-la-methode-de-stein/
SUMMARY:Nouveaux développements en statistique grâce à la méthode de St
 ein
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Christo
 phe Ley (Ghent University) : La méthode de Stein est un outil bien connu 
 en probabilités pour construire des bornes précises sur des distances pr
 obabilistes. Initialement proposée pour l’approximation gaussienne\, el
 le a par la suite été étendue à bon nombre de lois comme la loi de Poi
 sson\, binomiale\, exponentielle\, variance Gamma\, et bien d’autres. Ce
 s dernières années\, cette méthode probabiliste a aussi connu un réel 
 succès en statistique et machine learning\, et a permis des développemen
 ts théoriques et computationnels assez spectaculaires. Dans cet exposé\,
  je vais donner un aperçu sur ces développements\, avec un focus particu
 lier sur une nouvelle mesure de l’impact du choix de la prior distributi
 on en statistique bayésienne.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de probabilités et statistique virtuelle\, IECL\, Nancy\, F
 rance
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de probabilités et statistique virtuelle:geo:
 0,0
END:VEVENT
BEGIN:VEVENT
UID:67@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20210325T091500
DTEND;TZID=Europe/Paris:20210325T101500
DTSTAMP:20210322T191122Z
URL:https://iecl.univ-lorraine.fr/events/titre-a-venir-17/
SUMMARY:Rebondissements de mouvements browniens asymétriques
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Miguel 
 Martinez : \nDans cet exposé nous présenterons des résultats concernant
  les "rebonds" de deux mouvements browniens asymétriques (ou 'skew browni
 an motion') l'un sur l'autre. Nous verrons que dans une échelle de temps 
 adéquat\, la distance entre les deux processus se trouve être solution d
 'une équation différentielle stochastique à sauts dirigée par le proce
 ssus des excursions de l'un des deux mouvements\, tandis que les rebonds e
 ux-mêmes peuvent se décrire en faisant appel à la théorie des extensio
 ns markoviennes des processus auto-similaires. La fin de l'exposé sera co
 nsacrée à la présentation de certaines perspectives ouvertes par cette 
 étude.\n\n
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de probabilités et statistique virtuelle\, IECL\, Nancy\, F
 rance
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de probabilités et statistique virtuelle:geo:
 0,0
END:VEVENT
BEGIN:VEVENT
UID:19@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20210325T104500
DTEND;TZID=Europe/Paris:20210325T114500
DTSTAMP:20210318T073925Z
URL:https://iecl.univ-lorraine.fr/events/titre-a-venir-6/
SUMMARY:Mind2Mind: Transfer learning for GANs
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Yaël F
 régier : In this talk\, we will present a new approach to the problem of 
 transfer learning for GANs. It allows training deep neural networks with l
 imited computational resources in the specific context of generative model
 s. We prove rigorously\, within the framework of optimal transport\, a the
 orem that ensures the convergence of the learning of the transferred Wasse
 rstein GAN. It is joint work with Jean-Baptiste Gouray
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de probabilités et statistique virtuelle\, IECL\, Nancy\, F
 rance
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de probabilités et statistique virtuelle:geo:
 0,0
END:VEVENT
BEGIN:VEVENT
UID:68@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20210401T104500
DTEND;TZID=Europe/Paris:20210401T114500
DTSTAMP:20210325T093146Z
URL:https://iecl.univ-lorraine.fr/events/multilevel-picard-approximations-
 for-high-dimensional-semilinear-parabolic-partial-differential-equations/
SUMMARY:Multilevel Picard approximations for high-dimensional semilinear pa
 rabolic partial differential equations
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Thomas 
 Kruse (Justus Liebig University\, Giessen) : We present new approximation 
 methods for high-dimensional semilinear parabolic PDEs. A key idea of ou
 r methods is to combine multilevel approximations with Picard fixed-point 
 approximations. We prove in the case of semilinear heat equations with Lip
 schitz continuous nonlinearities that the computational effort of one of t
 he proposed methods grows polynomially both in the dimension and in the re
 ciprocal of the required accuracy. We illustrate the efficiency of the app
 roximation methods by means of numerical simulations. The talk is based on
  joint works with Weinan E\, Martin Hutzenthaler\, Arnulf Jentzen\, Tua
 n Nguyen and Philippe Von Wurstemberger.\n
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de probabilités et statistique virtuelle\, IECL\, Nancy\, F
 rance
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de probabilités et statistique virtuelle:geo:
 0,0
END:VEVENT
BEGIN:VEVENT
UID:69@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20210408T104500
DTEND;TZID=Europe/Paris:20210408T114500
DTSTAMP:20210331T152734Z
URL:https://iecl.univ-lorraine.fr/events/titre-a-venir-18/
SUMMARY:Factorisations de genre fixé d'un grand cycle
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Paul Th
 évenin (Uppsala University) : Une factorisation d'une permutation est une
  façon d'écrire cette permutation comme un produit de transpositions. L'
 ensemble des factorisations du n-cycle (12...n)\, particulièrement étudi
 é en raison notamment de ses liens avec la combinatoire algébrique\, est
  en bijection avec un ensemble de cartes à n sommets\, dont le genre est 
 donné par le nombre de transpositions de la factorisation. J'exposerai un
  algorithme inspiré de cette bijection et permettant de générer une fac
 torisation aléatoire uniforme du n-cycle dont la carte correspondante est
  de genre fixé.\n\nJe montrerai également comment cet algorithme permet 
 de décrire la limite\, en un certain sens\, d'une factorisation uniforme 
 de genre donné.\n\nTravail en collaboration avec Valentin Féray et Bapti
 ste Louf.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de probabilités et statistique virtuelle\, IECL\, Nancy\, F
 rance
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de probabilités et statistique virtuelle:geo:
 0,0
END:VEVENT
BEGIN:VEVENT
UID:511@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20210506T104500
DTEND;TZID=Europe/Paris:20210506T114500
DTSTAMP:20210430T095344Z
URL:https://iecl.univ-lorraine.fr/events/alex-watson/
SUMMARY:Strong laws for growth-fragmentation processes with bounded cell si
 ze
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Alex Wa
 tson (University College London) : A growth-fragmentation is a stochastic 
 process representing cells with continuously growing mass\, which experien
 ce sudden splitting events. Growth-fragmentations are used to model cell d
 ivision and protein polymerisation in biophysics. It is interesting to ask
  whether these processes converge toward an equilibrium\, in which the num
 ber of cells is growing exponentially and the distribution of cell sizes a
 pproaches some fixed asymptotic profile. In this work\, we study a process
  in which the growth and splitting of an individual cell is largely indepe
 ndent of its mass\, with the exception that the mass is bounded above\, so
  it cannot exceed a given constant. We give precise conditions to ensure t
 hat\, almost surely\, the process exhibits this equilibrium behaviour\, an
 d express the asymptotic profile in terms of an underlying Lévy process.\
 n\nThis is joint work with Emma Horton (Inria Bordeaux).
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:https://teams.microsoft.com/l/team/19%3add6f81af39af44dc878ce5a6a4
 4d9a2c%40thread.tacv2/conversations?groupId=55425192-2e79-495b-8d16-ce3e76
 fb6140&tenantId=158716cf-46b9-48ca-8c49-c7bb67e575f3
END:VEVENT
BEGIN:VEVENT
UID:79@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20210520T104500
DTEND;TZID=Europe/Paris:20210520T114500
DTSTAMP:20210510T192438Z
URL:https://iecl.univ-lorraine.fr/events/barbara-dembin/
SUMMARY:Principe de grande déviation pour les courants et le flot maximal 
 en percolation de premier passage
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Barbara
  Dembin (LPSM\, Paris) : Considérons la percolation de premier passage da
 ns le réseau renormalisé Z^d/n pour d&gt\;=2 : à chaque arête e\, on a
 ssocie une capacité aléatoire c(e)&gt\;=0 de telle sorte que la famille 
 (c(e))_e soit indépendante et identiquement distribuée selon une loi G. 
 On peut interpréter cette capacité comme un débit maximal\, i.e.\, la q
 uantité maximale d'eau pouvant traverser l'arête par unité de temps. Co
 nsidérons un domaine borné et connecté Ω de R^d et deux ensembles disj
 oints du bord de Ω : un part lequel l'eau peut entrer (la source) et un p
 art lequel l'eau peut sortir (le puits). Nous nous intéressons au flot ma
 ximal : la quantité maximale d'eau pouvant entrer dans Ω par unité de t
 emps. Un courant est une fonction sur les arêtes qui décrit la façon do
 nt l'eau circule dans Ω. Dans cet exposé\, nous présenterons un princip
 e de grande déviation pour les courants et nous en déduirons par un prin
 cipe de contraction un principe de grande déviation pour le flot maximal 
 dans Ω.\nTravail en collaboration avec Marie Théret.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:https://teams.microsoft.com/l/team/19%3add6f81af39af44dc878ce5a6a4
 4d9a2c%40thread.tacv2/conversations?groupId=55425192-2e79-495b-8d16-ce3e76
 fb6140&tenantId=158716cf-46b9-48ca-8c49-c7bb67e575f3
END:VEVENT
BEGIN:VEVENT
UID:80@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20210527T104500
DTEND;TZID=Europe/Paris:20210527T114500
DTSTAMP:20210520T122658Z
URL:https://iecl.univ-lorraine.fr/events/mouad-ramil/
SUMMARY:Langevin processes in bounded-in-position domains: application to q
 uasi-stationary distributions
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Mouad R
 amil (CERMICS\, Ecole des Ponts ParisTech) : Quasi-stationary distribution
 s can be seen as the first eigenvector associated with the generator of th
 e stochastic differential equation at hand\, on a domain with Dirichlet bo
 undary conditions (which corresponds to absorbing boundary conditions at t
 he level of the underlying stochastic processes). Many results on the quas
 i-stationary distribution hold for non degenerate stochastic dynamics\, wh
 ose associated generator is elliptic. The case of degenerate dynamics is l
 ess clear. In this work\, together with T. Lelièvre and J. Reygner (Ecole
  des Ponts\, France) we generalize well-known results on the probabilistic
  representation of solutions to parabolic equations on bounded domains to 
 the so-called kinetic Fokker-Planck equation on bounded domains in positio
 ns\, with absorbing boundary conditions. Furthermore\, a Harnack inequalit
 y\, as well as a maximum principle\, is provided for solutions to this kin
 etic Fokker-Planck equation\, as well as the existence of a smooth transit
 ion density for the associated absorbed Langevin dynamics. The continuity 
 of this transition density at the boundary is studied as well as the compa
 ctness\, in various functional spaces\, of the associated semigroup. This 
 work is a cornerstone to prove the consistency of some algorithms used to 
 simulate metastable trajectories of the Langevin dynamics\, for example th
 e Parallel Replica algorithm.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:https://teams.microsoft.com/l/team/19%3add6f81af39af44dc878ce5a6a4
 4d9a2c%40thread.tacv2/conversations?groupId=55425192-2e79-495b-8d16-ce3e76
 fb6140&tenantId=158716cf-46b9-48ca-8c49-c7bb67e575f3
END:VEVENT
BEGIN:VEVENT
UID:81@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20210603T104500
DTEND;TZID=Europe/Paris:20210603T114500
DTSTAMP:20210528T095755Z
URL:https://iecl.univ-lorraine.fr/events/pierre-yves-louis/
SUMMARY:Systèmes de processus de renforcement en interaction
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Pierre-
 Yves Louis (IMB\, Dijon) : Les modèles d'urnes sont utilisés dans de nom
 breuses applications et sont un exemple fondamental de processus stochasti
 ques de renforcement. En partant de ces modèles\, nous nous intéresseron
 s à plusieurs familles de systèmes (finis) de processus de renforcement.
  Différents résultats sur le comportement collectif en temps long seront
  présentés. La présence/absence de synchronisation sera discutée\, ain
 si que les vitesses de convergence en fonction de différents régimes de 
 paramètres. Cet exposé se fonde sur des travaux en collaboration avec I.
  Crimaldi\, P. Dai Pra\, I. Minelli et M. Mirebrahimi.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:https://teams.microsoft.com/l/team/19%3add6f81af39af44dc878ce5a6a4
 4d9a2c%40thread.tacv2/conversations?groupId=55425192-2e79-495b-8d16-ce3e76
 fb6140&tenantId=158716cf-46b9-48ca-8c49-c7bb67e575f3
END:VEVENT
BEGIN:VEVENT
UID:82@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20210610T104500
DTEND;TZID=Europe/Paris:20210610T114500
DTSTAMP:20210604T104410Z
URL:https://iecl.univ-lorraine.fr/events/elias-ventre/
SUMMARY:Reduction of a stochastic hybrid model of gene expression using Lar
 ge deviations theory
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Elias V
 entre (LBMC\, ENS Lyon) : Differentiation is the process whereby a cell ac
 quires a specific phenotype\, by differential gene expression as a functio
 n of time. This is thought to result from the dynamical functioning of an 
 underlying Gene Regulatory Network (GRN). The precise path from the stocha
 stic GRN behavior to the resulting cell state is still an open question. I
 n this presentation\, we detail a methodology to reduce a mechanistic mode
 l characterizing the evolution of a cell by a system of piecewise determin
 istic Markov processes (PDMP)\, to a discrete coarse-grained model on a li
 mited number of cell types\, defined as the basins of attraction of the de
 terministic limit. The transitions between the basins in the weak noise li
 mit can be determined by the unique solution of an Hamilton-Jacobi equatio
 n under a particular constraint\, which corresponds to the rate function a
 ssociated to a Large Deviations Principle for the PDMP. We develop a numer
 ical method for approximating the coarse-grained model parameters\, and sh
 ow its accuracy for a toggle-switch network. We deduce from the reduced mo
 del an analytical approximation of the stationary distribution of the PDMP
  system\, which appears as a Beta mixture.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:https://teams.microsoft.com/l/team/19%3add6f81af39af44dc878ce5a6a4
 4d9a2c%40thread.tacv2/conversations?groupId=55425192-2e79-495b-8d16-ce3e76
 fb6140&tenantId=158716cf-46b9-48ca-8c49-c7bb67e575f3
END:VEVENT
BEGIN:VEVENT
UID:83@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20210617T104500
DTEND;TZID=Europe/Paris:20210617T114500
DTSTAMP:20210608T164435Z
URL:https://iecl.univ-lorraine.fr/events/sonia-mazzucchi/
SUMMARY:High order heat-type equations and random walks on the complex plan
 e
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Sonia M
 azzucchi (Università di Trento\, Italie) : Télécharger le résumé (pdf
 )
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:https://teams.microsoft.com/l/team/19%3add6f81af39af44dc878ce5a6a4
 4d9a2c%40thread.tacv2/conversations?groupId=55425192-2e79-495b-8d16-ce3e76
 fb6140&tenantId=158716cf-46b9-48ca-8c49-c7bb67e575f3
END:VEVENT
BEGIN:VEVENT
UID:1710@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20210930T104500
DTEND;TZID=Europe/Paris:20210930T114500
DTSTAMP:20210826T155207Z
URL:https://iecl.univ-lorraine.fr/events/joseph-lam/
SUMMARY:Minimax optimality\, testing\, differential privacy
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Joseph 
 Lam (IECL\, Nancy) : This presentation is a summary of my PhD work. I focu
 s on the topic of hypothesis testing\, extensively studied in statistics a
 nd theoretical computer science.\n\nI start with presenting the classical 
 identity testing problem\, in which an independent sample set X ~ q is giv
 en and one would like to determine whether q=p for some fixed known p. Thi
 s problem is very related to that of estimating a distribution from a give
 n sample set. The study of testing is relevant\, because for the same fixe
 d sample size\, it is possible to test against a distribution up to a smal
 ler separation distance than what is possible in estimation. This will giv
 e me the opportunity to describe the minimax framework which proves the th
 eoretical optimality of statistical methods in the worst case.\n\nI will r
 efine the study of minimax identity testing by adding a local differential
  privacy condition and the interest will be in the quantitative effect of 
 ensuring privacy. The presentation will largely be on the topic of privacy
 \, because it bears similarities with ensuring fairness conditions.\n\nWe 
 will also shortly consider the neighboring problem of closeness testing\, 
 where the goal remains to determine whether p=q\, but only an independent 
 sample set Y ~ p is given instead of p directly. In this context\, we will
  go beyond a simple worst-case analysis and develop instance optimal resul
 ts instead. This will highlight the interplay between one-sample testing a
 nd two-sample testing\, the latter being a harder problem.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1726@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20211007T104500
DTEND;TZID=Europe/Paris:20211007T114500
DTSTAMP:20211001T115036Z
URL:https://iecl.univ-lorraine.fr/events/kelvin-rivera-lopez/
SUMMARY:Diffusions arising from the ordered Chinese Restaurant Process
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Kelvin 
 Rivera-Lopez (IECL\, Nancy) : In a recent paper\, Leonid Petrov showed tha
 t the up-down chains associated to the Chinese Restaurant Process (CRP) ha
 ve a scaling limit - namely\, a two-parameter family of diffusions that ex
 tend the one-parameter infinitely-many-neutral-alleles diffusions of Ethie
 r and Kurtz. There has since been considerable interest in constructing or
 dered analogues of Petrov's diffusions\, and it is conjectured that an ord
 ered analogue of the up-down chains will give rise to such an object. In t
 his talk\, I'll discuss my resolution of this conjecture (joint with Dougl
 as Rizzolo). Our approach is mainly inspired by Petrov's work\, and involv
 es using quasisymmetric functions to describe the transition operators.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle Döblin\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle Döblin:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1722@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20211014T104500
DTEND;TZID=Europe/Paris:20211014T114500
DTSTAMP:20211012T170223Z
URL:https://iecl.univ-lorraine.fr/events/eric-zeltz/
SUMMARY:Analyse et interprétation climatologique de l'évolution des temp
 ératures moyennes mondiales depuis 1880
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Eric Ze
 ltz : Je montre comment à partir d'une étude approfondie statistique et 
 probabiliste d'une base de données de températures moyennes mondiales\, 
 j'ai découvert des comportements climatologiques sans doute très diffici
 lement accessibles par les techniques usuelles utilisées en climatologie.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1727@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20211021T104500
DTEND;TZID=Europe/Paris:20211021T114500
DTSTAMP:20211012T161844Z
URL:https://iecl.univ-lorraine.fr/events/jean-berard/
SUMMARY:CFTP pour les automates cellulaires probabilistes uni-dimensionnels
  exponentiellement ergodiques
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Jean B
 érard (Université de Strasbourg) : Dans cet exposé\, on construit\, pou
 r tout automate cellulaire probabiliste uni-dimensionnel exponentiellement
  ergodique et possédant une propriété de taux positifs\, un flot CFTP (
 "coupling from the past") localement défini. Plusieurs conséquences de c
 ette construction sont discutées. (Travail exposé dans l'article arXiv:2
 106.07219).
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1723@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20211118T104500
DTEND;TZID=Europe/Paris:20211118T114500
DTSTAMP:20211115T232828Z
URL:https://iecl.univ-lorraine.fr/events/sebastien-darses/
SUMMARY:On probabilistic generalizations of the Nyman-Beurling criterion fo
 r the Zeta function
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Sébast
 ien Darses (Aix-Marseille Université) : One of the seemingly innocent ref
 ormulations of the terrifying Riemann Hypothesis (RH) is the Nyman-Beurlin
 g criterion: The indicator function of (0\,1) can be linearly approximated
  in a L^2 space by dilations of the fractional part function. Randomizing 
 these dilations generates new structures and criteria for RH\, regularizin
 g very intricate ones. One other possible nice feature is to consider poly
 nomials instead of Dirichlet polynomials for the approximations. How then 
 are the huge difficulties reallocated? The answers are quite surprising!\n
 \nThe talk will be very accessible\, especially for graduate students. \nJ
 oint work with F. Alouges and E. Hillion.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle Döblin\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle Döblin:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1724@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20211125T104500
DTEND;TZID=Europe/Paris:20211125T114500
DTSTAMP:20211104T104330Z
URL:https://iecl.univ-lorraine.fr/events/chiara-amorino/
SUMMARY:On the rate of estimation for the stationary distribution of stocha
 stic differential equations with and without jumps
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Chiara 
 Amorino (Université du Luxembourg) : In this talk\, we will discuss some 
 results on the estimation of the invariant density associated to a multiva
 riate diffusion X = (Xt)t≥0\, assuming that a continuous record of obser
 vations (Xt)0≤t≤T is available. We will see that\, when X = (Xt)t≥0 
 is the solution of a stochastic differential equation with Levy-type jumps
 \, it is possible to find the parametric convergence rate 1/T in the monod
 imensional case and log(T)/T when the dimension d is equal to 2. For d ≥
  3 we find the convergence rate (log(T)/T)γ\, where γ is an explicit exp
 onent depending on the dimension d and on β3\, the harmonic mean of the s
 moothness of the invariant density over the d directions after having remo
 ved β1 and β2\, which are the smallest. Moreover\, we obtain a lower bou
 nd on the L2-risk for pointwise estimation\, with the rate (1/T)γ. In ord
 er to fill the logarithmic gap we consider then X = (Xt)t≥0 as a solutio
 n to a continuous stochastic differential equation. One (surprising) findi
 ng is that the convergence rate depends on the fact that β2 &lt\; β3 or 
 β2 = β3. In particular\, we show that kernel density estimators achieve 
 the rate (log(T)/T)γ in the first case and (1/T)γ in the second. Finally
 \, we prove a minimax lower bound on the L2-risk for the pointwise estimat
 ion with the same rates (log(T)/T)γ or (1/T)γ\, depending on the value o
 f β2 and β3.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1734@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20211202T100000
DTEND;TZID=Europe/Paris:20211202T110000
DTSTAMP:20211125T143920Z
URL:https://iecl.univ-lorraine.fr/events/tony-bonnaire/
SUMMARY:Pattern extraction from point-cloud datasets and cosmological appli
 cations
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Tony Bo
 nnaire (Université Paris-Saclay) : Point-cloud datasets are ubiquitous in
  many science and non-science fields. These data are usually coming along 
 with unique patterns that some algorithms are meant to extract and that ar
 e linked with the underlying phenomenon that generated the data.\nIn this 
 presentation\, motivated by cosmological problematics\, we will focus on t
 wo kinds of spatially structured datasets. First\, clustered-type patterns
  in which the datapoints are separated in the input space into multiple gr
 oups. We will show that the unsupervised clustering procedure performed wi
 th a Gaussian Mixture Model can be formulated in terms of a statistical ph
 ysics optimisation problem. This formulation enables the unsupervised extr
 action of many key information about the dataset itself\, like the number 
 of clusters\, their size and how they are embedded in space\, particularly
  interesting for high-dimensional input spaces where visualisation is not 
 possible.\nOn the other hand\, we will study spatially continuous datasets
  assuming as standing on an underlying 1D structure that we aim to learn. 
 To this end\, we resort to a regularisation of the Gaussian Mixture Model 
 in which a spatial graph is used as a prior to approximate the underlying 
 1D structure. The overall graph is efficiently learnt by means of the Expe
 ctation-Maximisation algorithm with guaranteed convergence and comes toget
 her with the learning of the local width of the structure. We then illustr
 ate applications of the algorithm to model and identify the filamentary pa
 ttern drawn by the galaxy distribution of the Universe in cosmological dat
 asets.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle Döblin\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle Döblin:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1728@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20211209T104500
DTEND;TZID=Europe/Paris:20211209T114500
DTSTAMP:20211207T154841Z
URL:https://iecl.univ-lorraine.fr/events/aurelia-deshayes/
SUMMARY:Front du modèle FA1f en dimension 1 [REPORTÉ]
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Auréli
 a Deshayes (Université Paris-Est Créteil) : Dans cet exposé je présent
 erai un travail en collaboration avec Oriane Blondel et Cristina Toninelli
  où nous étudions le modèle FA1f en dimension 1. Il s’agit d’un sys
 tème de particules en interaction (plus précisément un modèle issu de 
 la physique statistique dit modèle cinétiquement contraint où chaque si
 te met à jour la valeur de son spin si une certaine contrainte locale est
  satisfaite\, ici c’est le fait d’avoir au moins un 0 dans ses voisins
 ). Dans ce travail\, nous prouvons\, sous certaines conditions\, une vites
 se linéaire\, et des fluctuations gaussiennes\, pour le front (i.e. le 0 
 le plus à gauche lorsque l’on part d’une configuration initiale avec 
 que des 1 à gauche de l’origine et un 0 en l’origine). Ce talk sera l
 ’occasion de présenter les techniques classiques utilisées dans les mo
 dèles de croissance aléatoire tels que le processus de contact et de par
 ler de méthode de couplage permettant de passer d’un modèle bien connu
  a un modèle plus complexe (en particulier non attractif).
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1725@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20211216T104500
DTEND;TZID=Europe/Paris:20211216T114500
DTSTAMP:20211210T183646Z
URL:https://iecl.univ-lorraine.fr/events/remi-catellier/
SUMMARY:Pathwise regularization of the stochastic heat equation with multip
 licative noise through irregular perturbation
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Rémi C
 atellier (Université Côte d'Azur) : Existence and uniqueness of solution
 s to the stochastic heat equation with multiplicative spatial noise is stu
 died. In the spirit of pathwise regularization by noise\, we show that a p
 erturbation by a sufficiently irregular continuous path establish wellpose
 dness of such equations\, even when the drift and diffusion coefficients a
 re given as generalized functions or distributions. In addition we prove r
 egularity of the averaged field associated to a Lévy fractional stable mo
 tion\, and use this as an example of a perturbation regularizing the multi
 plicative stochastic heat equation.\n\nJoint work With Fabian Harang
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1825@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20220106T104500
DTEND;TZID=Europe/Paris:20220106T114500
DTSTAMP:20220104T162828Z
URL:https://iecl.univ-lorraine.fr/events/oliver-tough/
SUMMARY:Stochastic approximation of the paths of killed Markov processes co
 nditioned on survival
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Oliver 
 Tough (Université de Neuchâtel) : Reinforced processes are known to prov
 ide a stochastic approximation for the quasi-stationary distributions of k
 illed Markov processes. We show how the construction may be adapted to pro
 vide a stochastic approximation of the paths of killed Markov processes co
 nditioned on survival. Whilst rigorous results are restricted to time bein
 g discrete and the state space finite\, the strategy employed should be ex
 tendable to a general setting in the future.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1733@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20220113T104500
DTEND;TZID=Europe/Paris:20220113T114500
DTSTAMP:20220106T224412Z
URL:https://iecl.univ-lorraine.fr/events/adrien-richou/
SUMMARY:À propos de l'espérance conditionnelle contrainte dans un domaine
  non convexe
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Adrien 
 Richou (Université de Bordeaux) : Je présenterai dans cet exposé des r
 ésultats nouveaux sur l'existence et l'unicité de solution pour des EDSR
 s réfléchies dans des domaines non convexes supposés "faiblement étoil
 és". Notons que le cas particulier des EDSRs de générateur nul\, à sav
 oir l'espérance conditionnelle pour la filtration brownienne\, est déjà
  un cas d'étude intéressant et permet de définir une notion de moyenne 
 contrainte à prendre ses valeurs dans un ensemble non convexe. En particu
 lier\, on établit des résultats d'existence et d'unicité dans un cadre 
 markovien avec une condition terminale et un générateur réguliers\, mai
 s également dans un cadre général sous une hypothèse de petitesse sur 
 les paramètres de l'EDSR. C'est un travail en commun avec Jean-François 
 Chassagneux (Université de Paris) et Sergey Nadtochiy (Illinois Institute
  of Technology).
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1846@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20220120T104500
DTEND;TZID=Europe/Paris:20220120T114500
DTSTAMP:20220114T160900Z
URL:https://iecl.univ-lorraine.fr/events/nicolas-chenavier/
SUMMARY:Durées de vie extrémales en analyse topologique des données
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Nicolas
  Chenavier (Université du Littoral Côte d'Opale) : L'un des principes de
  l'analyse topologique des données est d'étudier un ensemble de données
 \, représentées sous forme d'un nuage de points\, à partir d'outils top
 ologiques. Un concept de base est celui de l'homologie persistante. Cette 
 dernière mesure les naissances et les morts de diverses caractéristiques
  topologiques\, telles que les boucles et les cavités\, lorsque l'on fait
  grossir des boules en chaque point d'un processus de Poisson (on parle de
  modèle Booléen). Dans cet exposé\, nous nous intéressons aux durées 
 de vie extrémales pour de telles caractéristiques. Nous étudions d'abor
 d le cas particulier des cavités et donnons l'ordre de grandeur du maximu
 m (resp. du minimum) de leurs durées de vie. Une approximation poissonien
 ne du nombre d'excédents est également établie. Nous étendons ensuite 
 l'étude à des caractéristiques quelconques pour les complexes simplicia
 ux de Cech et de Vietoris-Rips. Travail joint avec C. Hirsch.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1840@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20220127T104500
DTEND;TZID=Europe/Paris:20220127T114500
DTSTAMP:20220118T134102Z
URL:https://iecl.univ-lorraine.fr/events/julien-jacques/
SUMMARY:Functional data clustering with outlier detection
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Julien 
 Jacques (Université Lumière Lyon 2) : With the emergence of numerical se
 nsors in many aspects of every-day life\, there is an increasing need in a
 nalyzing high frequency data\, which can be seen as discrete observation o
 f functional data.\nThe presentation will focus on the clustering of such 
 functional data\, in order to ease their modeling and understanding. To th
 is end\, a novel clustering technique for multivariate functional data is 
 presented.\nThis method is based on a functional latent mixture model whic
 h fits the data in group-specific functional subspaces through a multivari
 ate functional principal component analysis. \nIn such clustering analysis
 \, the presence of outliers can confuse the notion of cluster.\nConsequent
 ly\, a contaminated version of the previous mixture model is proposed. Thi
 s model both clusters the multivariate functional data into homogeneous gr
 oups and detects outliers. The main advantage of this procedure over its c
 ompetitors is that it does not require us to specify the proportion of out
 liers.\nModel inference is performed through an Expectation-Conditional Ma
 ximization algorithm\, and the BIC criterion is used to select the number 
 of clusters. Numerical experiments on simulated data demonstrate the high 
 performance achieved by the inference algorithm. In particular\, the propo
 sed model outperforms competitors. Its application on the real data which 
 motivated this study allows us to correctly detect abnormal behaviors.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle Döblin\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle Döblin:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1769@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20220203T104500
DTEND;TZID=Europe/Paris:20220203T114500
DTSTAMP:20220126T132108Z
URL:https://iecl.univ-lorraine.fr/events/david-coupier/
SUMMARY:La forêt IDLA
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 David C
 oupier (Institut Mines Télécom Nord Europe) : Le modèle IDLA (Internal 
 Diffusion Limited Aggregation) est un modèle de croissance aléatoire sur
  la grille Zd introduit dans les années 90 et permettant de modéliser l'
 évolution d'un bassin de culture de cellules\, la croissance de zones urb
 aines ou encore la propagation d'un fluide visqueux. C'est une suite d'ens
 embles aléatoires (An)n définie comme suit : A0 = {0} et\, étant donné
  An\, on lance une marche aléatoire simple depuis l'origine de Zd. Le som
 met z par lequel la marche sort de l'agrégat An est ajouté pour obtenir 
 An+1 : An+1 = An U {z}. Un arbre aléatoire se cache derrière la suite de
 s agrégats (An)n... Afin d'étudier la géométrie de cet arbre\, nous av
 ons défini en 2020 un graphe aléatoire auxiliaire\, baptisé la forêt d
 irigée IDLA. Ce nouvel objet possède d'intéressantes propriétés et de
 s conjectures excitantes qui seront abordées dans cet exposé. Travail en
  collaboration avec N. Chenavier (ULCO) et A. Rousselle (Dijon)
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1867@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20220224T104500
DTEND;TZID=Europe/Paris:20220224T114500
DTSTAMP:20220211T113414Z
URL:https://iecl.univ-lorraine.fr/events/yvain-bruned/
SUMMARY:Renormalisation locale pour les EDPS singulières
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Yvain B
 runed (University of Edinburgh) : Dans cet exposé\, on présentera les ou
 tils des structures de régularité pour traiter les EDP stochastiques sin
 gulières qui ne sont pas invariantes par translation. On décrira en part
 iculier l'équation renormalisée pour une très large classe de schémas 
 de renormalisation dépendant de l'espace-temps. Cette approche est basée
  sur des renormalisations locales qui agissent directement au niveau de l'
 équation. L'exposé sera basé sur un travail en collaboration avec Isma
 ël Bailleul.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1898@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20220303T104500
DTEND;TZID=Europe/Paris:20220303T114500
DTSTAMP:20220224T165309Z
URL:https://iecl.univ-lorraine.fr/events/mazyar-ghani-varzaneh/
SUMMARY:Dynamical properties of rough delay equations
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Mazyar 
 Ghani Varzaneh (Technische Universität Berlin) : This talk aims to incorp
 orate two subjects for developing a framework for studying the long-time b
 ehavior solution of singular delay equations. Singular delay equations fai
 l to induce the flow property. Accordingly\, for a long time\, many people
  have believed it is not possible to apply the idea of random dynamical sy
 stems to this family of equations.\nIn this talk\, we claim\, is possible.
  The main trick is to regard the solution in the language of the Rough pat
 h and then construct the flow property in a bundle-like family of Banach s
 paces. The main challenge here is to prove the Multiplicative Ergodic Thro
 em in this new framework. After proving this crucial theorem\, we can gene
 rate the Lyapunov exponents. These exponents can be regarded as a generali
 zation of eigenvalues. We then apply these theorems to prove the invariant
  manifolds in our setting. The main tools here are the rough path theory a
 nd random dynamical systems.\nThis talk is based on my doctoral thesis. I 
 recently have defended my thesis in February.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:https://teams.microsoft.com/l/meetup-join/19%3add6f81af39af44dc878
 ce5a6a44d9a2c%40thread.tacv2/1610639873485?context=%7b%22Tid%22%3a%2215871
 6cf-46b9-48ca-8c49-c7bb67e575f3%22%2c%22Oid%22%3a%22bc9c1296-f22e-418e-812
 b-e92cbd1c7362%22%7d
END:VEVENT
BEGIN:VEVENT
UID:1770@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20220310T104500
DTEND;TZID=Europe/Paris:20220310T114500
DTSTAMP:20220301T161210Z
URL:https://iecl.univ-lorraine.fr/events/thomas-opitz/
SUMMARY:Les modèles de processus ponctuel spatiotemporels avec marques ext
 rêmes : une application aux feux de forêts en France
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Thomas 
 Opitz (INRAE Avignon) : Accurate spatiotemporal modeling of conditions lea
 ding to moderate and large wildfires provides better understanding of mech
 anisms driving fire-prone ecosystems and improves risk management. We here
  develop a joint model for the occurrence intensity and the wildfire size 
 distribution by combining extreme-value theory and point processes within 
 a novel Bayesian hierarchical model\, and use it to study daily summer wil
 dfire data for the French Mediterranean basin during 1995-2018. The occurr
 ence component models wildfire ignitions as a spatiotemporal log-Gaussian 
 Cox process. Burnt areas are numerical marks attached to points and are co
 nsidered as extreme if they exceed a high threshold. The size component is
  a two-component mixture varying in space and time that jointly models mod
 erate and extreme fires. We capture non-linear influence of covariates (Fi
 re Weather Index\, forest cover) through component-specific smooth functio
 ns\, which may vary with season. We propose estimating shared random effec
 ts between model components to reveal and interpret common drivers of diff
 erent aspects of wildfire activity. This leads to increased parsimony and 
 reduced estimation uncertainty with better predictions. Fast approximate (
 but accurate) Bayesian estimation is carried out in the framework of the i
 ntegrated nested Laplace approximation. Our methodology provides a holisti
 c approach to explaining and predicting the drivers of wildfire activity a
 nd associated uncertainties.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1771@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20220317T104500
DTEND;TZID=Europe/Paris:20220317T114500
DTSTAMP:20220309T181049Z
URL:https://iecl.univ-lorraine.fr/events/vincent-duval/
SUMMARY:Une méthode « sans grille » pour la reconstruction d'images
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Vincent
  Duval (INRIA Paris) : Ces dernières années\, les méthodes de reconstru
 ction avec a priori de parcimonie (LASSO\, Basis Pursuit)\, très utilisé
 es en statistiques comme en traitement d'images\, ont été adaptées pour
  opérer sur un domaine continu (Beurling Minimal extrapolation\, Beurling
 -LASSO...): on reconstruit alors une somme de masses de Dirac plutôt qu'u
 n vecteur parcimonieux.\nLe fait de travailler sur un domaine continu appo
 rte de nombreux avantages: absence de grille de reconstruction et des arte
 facts de discrétisation associés\, analyse plus simple\, et algorithmes 
 tirant parti de la structure lisse du problème.\n\nDans cet exposé\, nou
 s nous proposons d'étendre cette démarche à la reconstruction d'objets 
 plus complexes: plutôt que des sources ponctuelles\, on veut reconstruire
  des images constantes par morceaux à l'aide de la régularisation par va
 riation totale du gradient (comme dans les travaux de Rudin\, Osher et Fat
 emi).\nNous montrons qu'en étudiant la boule unité associée\, on peut d
 écrire la structure des minimiseurs et définir un algorithme de type Fra
 nk-Wolfe « sans grille » pour la résolution du problème.\nL'avantage d
 'une telle méthode est la préservation des bords et l'isotropie des solu
 tions.\n\nIl s'agit d'un travail commun avec Romain Petit et Yohann De Cas
 tro.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1812@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20220324T104500
DTEND;TZID=Europe/Paris:20220324T114500
DTSTAMP:20220304T194916Z
URL:https://iecl.univ-lorraine.fr/events/sebastien-martineau/
SUMMARY:Percolation surcritique sur les graphes à croissance polynomiale
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Sébast
 ien Martineau (LPSM\, Paris) : La percolation consiste à partir d'un grap
 he raisonnable G\, d'un paramètre p dans [0\,1] et à conserver chaque ar
 ête indépendamment avec probabilité p\, effaçant toutes les autres. On
  s'intéresse alors aux composantes connexes du graphe ainsi formé (ces c
 omposantes sont appelées amas ou clusters). Par exemple\, existe-t-il un 
 cluster infini ?\n\nIl existe un paramètre critique p_c\, qui dépend du 
 graphe\, tel que :\n- pour tout p &lt\; p_c\, il n'y ait presque sûrement
  aucun cluster infini\,\n- pour tout p &gt\; p_c\, il existe presque sûre
 ment (au moins) un cluster infini.\n\nLe régime sous-critique (p &lt\; p_
 c) est bien compris\, et ce pour des graphes généraux. Le régime critiq
 ue (p = p_c) est considérablement plus difficile : il fait l'objet de gra
 nds théorèmes et conjectures. C'est au régime restant\, le surcritique 
 (p &gt\; p_c)\, que sera dédié cet exposé. Ce régime est plus difficil
 e que le sous-critique mais moins ardu que le régime critique.\n\nContrai
 rement au régime sous-critique\, le régime surcritique est\, en un certa
 in sens qu'on précisera\, sensible à la géométrie du graphe de départ
 . Il est donc raisonnable de se restreindre à certaines classes de graphe
 s définies par des hypothèses géométriques. On verra qu'en se restreig
 nant aux graphes dits "à croissance polynomiale"\, il est possible d'obte
 nir une compréhension fine du régime surcritique. Cela permet de retrouv
 er par des techniques nouvelles des résultats déjà connus sur le résea
 u cubique (Grimmett–Marstrand...)\, ainsi que de couvrir toute une gamme
  de graphes intéressants (discrétisations anisotropes de Z^d\, graphes d
 e Cayley de groupes nilpotents).\n\nCet exposé porte sur des travaux réa
 lisés en collaboration avec Daniel Contreras et Vincent Tassion.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1922@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20220407T104500
DTEND;TZID=Europe/Paris:20220407T114500
DTSTAMP:20220401T160444Z
URL:https://iecl.univ-lorraine.fr/events/dylan-dronnier/
SUMMARY:Modèles d’épidémie en dimension infinie et stratégie de vacci
 nation optimale
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Dylan D
 ronnier (Université de Neuchâtel) : Dans une population homogène\, le n
 ombre de reproduction de base\, noté R0\, est défini comme le nombre moy
 en de cas directement générés par une personne contagieuse quand tous l
 es autres individus sont sains et sensibles à l’infection. Ce nombre jo
 ue un rôle fondamental en épidémiologie puiqu’il constitue un seuil q
 ui détermine si l’épidémie va finir par disparaître (cas R0 ≤ 1) o
 u\, au contraire\, devenir endémique (cas R0 &gt\; 1).\nImaginons désorm
 ais qu’une proportion 1 − 1/R0 de la population est immunisée (en ét
 ant vaccinée par exemple). Le nombre moyen d’individus qu’infecte la 
 personne contagieuse est alors divisé par R0. On en déduit que le nouvea
 u nombre de reproduction\, qualifié d’effectif dans ce cas\, est égal 
 à 1 : l’épidémie finira par disparaître grâce au phénomène d’im
 munité grégaire. Le nombre 1 − 1/R0\, appelé seuil d’immunité de g
 roupe\, est souvent utilisé pour evaluer l’efficacité d’une politiqu
 e de vaccination par les autorités sanitaires.\nQuand les contacts dans l
 a population ne sont plus homogènes\, le nombre de reproduction de base e
 st défini comme le nombre de cas directement générés par une personne 
 infectée typique quand tous les autres individus sont sains et sensibles 
 à l’infection. Le seuil d’immunité collective 1 − 1/R0 reste encor
 e valide quand on vaccine la population uniformément. Il est cependant na
 turel de se demander si l’on ne pourrait pas abaisser ce seuil en ciblan
 t certains groupes dans la population.\nL’objectif de cette présentatio
 n est de proposer une formalisation mathématique de ce problème. Pour mo
 déliser les contacts dans la population\, j’utiliserai des objets issus
  de la théorie des limites des grands graphes. Dans la première partie d
 e l’exposé\, je présenterai un modèle hétérogène de type SIS (Susc
 eptible → Infecté → Susceptible) avec vaccination que nous avons intr
 oduit récemment. Ce modèle servira de base pour définir les stratégies
  optimales de vaccination\, montrer leur existence et étudier leurs propr
 iétés de stabilité. Enfin\, je donnerai une série d’exemples où les
  solutions du problème de vaccination optimale peuvent être exprimées d
 e manière analytique.\n\n→ Version PDF avec références
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1879@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20220428T104500
DTEND;TZID=Europe/Paris:20220428T114500
DTSTAMP:20220421T111810Z
URL:https://iecl.univ-lorraine.fr/events/patricia-reynaud-bouret/
SUMMARY:Simulation de grands réseaux de neurones
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Patrici
 a Reynaud-Bouret (Université Côte d'Azur) : Après avoir présenté les 
 défis numériques actuels pour atteindre des tailles de réseaux de l'ord
 re du cerveau humain\, j'expliquerai pourquoi les processus de Hawkes peuv
 ent être un bon modèle pour passer à l'échelle. J'expliquerai comment 
 les algorithmes classiques peuvent être renouvelés pour y arriver. Une d
 es approches les plus innovantes est basée sur un résultat probabiliste 
 fort : la décomposition de Kalikow\, qui permet de tirer au hasard les d
 épendances spatiales et dans le passé et que nous avons redéfini en tem
 ps continu. Elle permet en particulier de simuler un neurone immergé dans
  un réseau infini sans avoir à simuler le réseau infini. Ce travail est
  effectué en collaboration avec Eva Löcherbach (mathématicienne\, Paris
  I)\, Alexandre Muzy (informaticien\, Université Côte d'Azur) ainsi que 
 nos étudiants : Cyrille Mascart\, Tien Cuong Phi et Paul Gresland.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1885@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20220505T104500
DTEND;TZID=Europe/Paris:20220505T114500
DTSTAMP:20220124T134604Z
URL:https://iecl.univ-lorraine.fr/events/aurelia-deshayes-2/
SUMMARY:Front du modèle FA1f en dimension 1
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Aurelia
  Deshayes (Université Paris-Est Créteil) : Dans cet exposé je présente
 rai un travail en collaboration avec Oriane Blondel et Cristina Toninelli 
 où nous étudions le modèle FA1f en dimension 1. Il s’agit d’un syst
 ème de particules en interaction (plus précisément un modèle issu de l
 a physique statistique dit modèle cinétiquement contraint où chaque sit
 e met à jour la valeur de son spin si une certaine contrainte locale est 
 satisfaite\, ici c’est le fait d’avoir au moins un 0 dans ses voisins)
 . Dans ce travail\, nous prouvons\, sous certaines conditions\, une vitess
 e linéaire\, et des fluctuations gaussiennes\, pour le front (i.e. le 0 l
 e plus à gauche lorsque l’on part d’une configuration initiale avec q
 ue des 1 à gauche de l’origine et un 0 en l’origine). Ce talk sera l
 ’occasion de présenter les techniques classiques utilisées dans les mo
 dèles de croissance aléatoire tels que le processus de contact et de par
 ler de méthode de couplage permettant de passer d’un modèle bien connu
  a un modèle plus complexe (en particulier non attractif).
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1883@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20220512T104500
DTEND;TZID=Europe/Paris:20220512T114500
DTSTAMP:20220506T163737Z
URL:https://iecl.univ-lorraine.fr/events/florent-koechlin/
SUMMARY:Réductions d’arbres aléatoires d’expressions en présence d
 ’un élément absorbant
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Florent
  Koechlin (Loria) : En informatique\, les expressions aléatoires sont cou
 ramment utilisées pour analyser des algorithmes\, que ce soit pour étudi
 er leur complexité en moyenne\, ou pour générer des benchmarks pour les
  tester expérimentalement. Généralement\, ces approches considèrent le
 s expressions en entrée comme des arbres purement syntaxiques\, et font a
 bstraction de leur sémantique\, c’est-à-dire de l’objet mathématiqu
 e représenté par l’expression. \n\nPourtant\, deux expressions différ
 entes peuvent être équivalentes (par exemple « 0*(x+y) » et « 0 »  r
 eprésentent la même expression\, l’expression  nulle). Ces phénomène
 s de redondances remettent-ils en question la pertinence de ces analyses e
 t ces tests qui ne tiennent pas compte de la sémantique des expressions ?
 \n\nJe présenterai comment la distribution uniforme sur les arbres syntax
 iques d'expressions devient complètement dégénérée lorsqu’on commen
 ce à prendre en compte leur sémantique\, dans le cas très simple mais c
 ourant où il existe un élément absorbant. Si le temps le permet\, j’e
 xpliquerai pourquoi la distribution ABR laisse plus d’espoirs.\n\nIl s
 ’agit d’un travail effectué pendant ma thèse\, en commun avec Cyril 
 Nicaud et Pablo Rotondo.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1855@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20220519T104500
DTEND;TZID=Europe/Paris:20220519T114500
DTSTAMP:20220527T170327Z
URL:https://iecl.univ-lorraine.fr/events/guillaume-chapuy/
SUMMARY:Partitions aléatoires\, cartes de grand genre et marche aléatoire
  sur les permutations
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Guillau
 me Chapuy (Université de Paris) : Je m'intéresserai à un modèle de par
 titions d'entiers aléatoires qui est une déformation à un paramètre de
  la très classique mesure de Plancherel du groupe symétrique. Cette déf
 ormation\, qui a une définition combinatoire explicite\, a sa source dans
  la théorie des nombres de Hurwitz\, qui comptent certaines familles de c
 artes plongées sur des surfaces. La déformation que nous étudions inter
 vient naturellement lorsque l'on s'intéresse à des nombres de Hurwitz (o
 u des cartes) de très grand genre\, un problème qui se formule égalemen
 t dans le langage de la marche aléatoire sur le groupe des permutations e
 ngendré par les transpositions. Nous exhibons un phénomène de forme lim
 ite d'un type nouveau pour ces partitions\, qui a des conséquences pour l
 'énumération des cartes et pour la marche. La démonstration utilise une
  méthode dite "entropique" qui mélange un peu de calcul des variations 
 à beaucoup d'estimées combinatoires.\nL'exposé sera introductif\, sans 
 pré-requis\, avec de jolies images.\nTravail en commun avec Baptiste Louf
  et Harriet Walsh.\n\nThis project has received funding from the European 
 Research Council (ERC) under the European Union’s Horizon 2020 research 
 and innovation programme (grant agreement No. ERC-2016-STG 716083 “Combi
 Top”).
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1917@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20220609T104500
DTEND;TZID=Europe/Paris:20220609T114500
DTSTAMP:20220509T094538Z
URL:https://iecl.univ-lorraine.fr/events/jerome-casse/
SUMMARY:Le mouvement brownien itéré ad libitum n'est pas le pseudo-arc
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Jérôm
 e Casse (Université Paris-Saclay) : À partir d'une suite de mouvements b
 rowniens bilatères indépendants\, Kiss et Solecki ont construit un conti
 nuum (un espace métrique connexe\, compact et non vide) aléatoire. Ils o
 nt montré que ce continuum est indécomposable p.s. Avec Nicolas Curien\,
  nous avons montré qu'il n'est pas héréditairement indécomposable p.s.
 \, et que ce n'est donc pas le pseudo-arc.\n\nDans cet exposé\, j'expliqu
 erai l'ensemble des termes précédents\, la construction de ce continuum 
 aléatoire et je vous expliquerai pourquoi il est indécomposable\, mais p
 as héréditairement indécomposable.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1758@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20220616T104500
DTEND;TZID=Europe/Paris:20220616T114500
DTSTAMP:20220614T140517Z
URL:https://iecl.univ-lorraine.fr/events/claire-boyer/
SUMMARY:Sampling Rates for ℓ1-Synthesis
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Claire 
 Boyer (Sorbonne Université) : ...ou « Combien de projections sous-gauss
 iennes doit-on faire pour reconstruire un objet parcimonieux dans un dicti
 onnaire redondant ? »\n\nThis work investigates the problem of signal re
 covery from undersampled noisy sub-Gaussian measurements under the assumpt
 ion of a synthesis-based sparsity model. Solving the l1-synthesis basis pu
 rsuit allows to simultaneously estimate a coefficient representation as we
 ll as the sought-for signal. However\, due to linear dependencies within r
 edundant dictionary atoms it might be impossible to identify a specific re
 presentation vector\, although the actual signal is still successfully rec
 overed. We study both estimation problems from a non-uniform\, signal-depe
 ndent perspective. By utilizing results from linear inverse problems and c
 onvex geometry\, we identify the sampling rate describing the phase transi
 tion of both formulations\, and propose a "tight" estimated upper-bound.\n
 \nThis is a joint work with Maximilian März (TU Berlin)\, Jonas Kahn and 
 Pierre Weiss (CNRS\, Toulouse).
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle Döblin\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle Döblin:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1994@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20221006T104500
DTEND;TZID=Europe/Paris:20221006T114500
DTSTAMP:20220928T062005Z
URL:https://iecl.univ-lorraine.fr/events/yvain-bruned-2/
SUMMARY:Approximation d'EDP dispersives en présence d'un aléa de faible r
 égularité
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Yvain B
 runed (Université de Lorraine) : Dans cet exposé\, on introduit une nouv
 elle classe de schémas numériques qui permettent des approximations de f
 aible régularité du second moment de la solution d'une EDP dispersive av
 ec des données initiales aléatoires. Cette quantité joue un rôle impor
 tant en physique\, en particulier dans l'étude de la turbulence des ondes
  où il faut adopter une approche statistique afin d'obtenir une compréhe
 nsion approfondie du comportement générique à long terme des solutions 
 aux équations dispersives. Nos schémas utilisent une discrétisation bas
 ée sur la résonance après avoir appliqué le théorème de Wick qui pro
 duit des diagrammes de Feynman. Pour écrire ces schémas\, on introduit d
 es forêts décorées appariées qui sont deux arbres décorés dont les d
 écorations sur les feuilles viennent par paires. La construction du sché
 ma s'inspire du traitement des équations aux dérivées partielles stocha
 stiques singulières via les structures de régularité. Il s'agit d'un tr
 avail conjoint avec Yvonne Alama Bronsard et Katharina Schratz.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1995@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20221013T104500
DTEND;TZID=Europe/Paris:20221013T114500
DTSTAMP:20221006T215950Z
URL:https://iecl.univ-lorraine.fr/events/yoann-offret/
SUMMARY:Marches aléatoires maximales entropiques (MAMEs) et limites d'éch
 elles
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Yoann O
 ffret (Université de Bourgogne) : Les MAMEs sur des graphes sont les MAs 
 qui maximisent l'entropie trajectoriellement. Leurs constructions nécessi
 tent une connaissance globale du graphe dont le rayon spectral et vecteur 
 propre positif associé. A contrario\, les MA simples peuvent être vues c
 omme celles maximisant l'entropie localement.\n\nCes MAs ont été introdu
 ites il y a une dizaine d'années par des physiciens et des informaticiens
 . Elles ont par exemple de meilleures propriétés diffusives dans les ré
 seaux réguliers et de fortes propriétés de localisations dans les milie
 ux irréguliers. Elles ont déjà trouvé de nombreuses applications dans 
 le traitement d'images ou la prédiction de liens dans un graphe notamment
 .\n\nJe présenterai quelques exemples\, notamment des MAMEs sur des graph
 es infinis et des processus d'exclusions maximaux entropiques\, et je parl
 erai de certaines limites d'échelles de ces processus (Bessel 3\, Mouveme
 nt Brownien de Dyson...).
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1996@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20221020T104500
DTEND;TZID=Europe/Paris:20221020T114500
DTSTAMP:20221018T074335Z
URL:https://iecl.univ-lorraine.fr/events/josue-corujo/
SUMMARY:A dynamical approach to spanning and surplus edges of random graphs
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Josué 
 Corujo (Université de Strasbourg) : During this talk\, we will review som
 e recent advances in the multiplicative coalescent theory and its link to 
 random graphs. The multiplicative coalescent dynamic naturally emerges whe
 n one regards the evolution of the connected components in a graph-valued 
 Markov process. We will mainly focus on the breadth-first walk introduced 
 by V. Limic (2019)\, a Lévy-type process encoding a random forest whose c
 omponents (trees) are a representation of the multiplicative coalescent. W
 e will then focus on the extension of this construction to account for the
  surplus edges data\, in addition to the spanning edge data. We will prese
 nt two different graph representations of the multiplicative coalescent\, 
 with different advantages and drawbacks\, that are discussed in detail. In
  particular\, we will show how to recover a realization of the random grap
 h at a fixed time\, and also as a process when the time parameter evolves.
  We will also discuss the use of these results to understand the scaling l
 imits of near-critical random graphs in the domain of attraction of genera
 l eternal multiplicative coalescent.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1997@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20221110T104500
DTEND;TZID=Europe/Paris:20221110T114500
DTSTAMP:20221103T141357Z
URL:https://iecl.univ-lorraine.fr/events/laetitia-colombani/
SUMMARY:Some asymptotic properties of inhibitive Hawkes process
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Laetiti
 a Colombani (Universität Bern) : Hawkes process was introduced by Hawkes 
 in 1971 and are widely used in many applications (earthquakes\, neurons\, 
 social network\, finance\, etc.). This jump process has an intensity which
  depends on the past. Linear "self-exciting" Hawkes process has been parti
 cularly studied and some asymptotic results are well-known.\nDuring my PhD
 \, with Manon Costa and Patrick Cattiaux\, I considered non-linear Hawkes 
 processes\, which can model self-inhibition and self-excitation. We proved
  asymptotic properties (law of large numbers\, CLT\, large deviations)\, b
 y considering a new point of view for this process: the renewal structure 
 of some Hawkes process leads to a comparison with cumulative processes.\nI
 n this talk\, I'll introduce Hawkes processes and cumulative processes. By
  exhibiting their link\, I'll give an idea of the approach we use to prove
  asymptotic properties.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1998@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20221117T104500
DTEND;TZID=Europe/Paris:20221117T114500
DTSTAMP:20221110T180709Z
URL:https://iecl.univ-lorraine.fr/events/gaspar-massiot/
SUMMARY:Monitoring the risk of Legionella infection using general Bayesian 
 network updated from temporal measurements in agricultural irrigation with
  reclaimed wastewater
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Gaspar 
 Massiot (AgroParisTech Nancy) : General Bayesian Networks (GBNs) extend Ba
 yesian networks to the modeling of continuous links in the data.\nI will d
 emonstrate the implementation of the GBNs in the context of risk monitorin
 g for Legionella infection from the use of reclaimed wastewater to irrigat
 e agricultural plots.\nI will also discuss the use of these networks to ev
 aluate hypothetical scenarios of how failures of the system propagate in t
 he model.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:1999@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20221124T104500
DTEND;TZID=Europe/Paris:20221124T114500
DTSTAMP:20221114T212807Z
URL:https://iecl.univ-lorraine.fr/events/giovanni-conforti/
SUMMARY:Théorèmes de turnpike en contrôle stochastique
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Giovann
 i Conforti (École Polytechnique) : Nous nous intéressons au comportement
  en temps long des processus de Markov obtenus comme solutions optimales d
 e problèmes d’optimisation stochastique\, comme par exemple des problè
 mes de contrôle stochastique ou des problèmes de transport optimal stoch
 astique. Dans ce contexte\, le générateur du processus n’est pas connu
  en forme explicite et depend de la solution d’une  EDP non linéaire\, 
 typiquement une équation d'Hamilton-Jacobi-Bellman. Le but de cet exposé
  est d’expliquer comment on peut définir une notion de mesure invariant
 e\, qu’on appelle turnpike dans ce cadre\, et d’illustrer les idées d
 e base d’une technique par couplage qui permet d’obtenir des résultat
 s de convergence exponentielle vers le turnpike. Dans un deuxième temps\,
  la question der comment éteindre ces notions et résultats au contrôle 
 McKean-Vlasov sera aussi abordée.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2000@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20221201T104500
DTEND;TZID=Europe/Paris:20221201T114500
DTSTAMP:20221128T101234Z
URL:https://iecl.univ-lorraine.fr/events/ester-mariucci/
SUMMARY:Nonparametric estimation of the Lévy density
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Ester M
 ariucci (Université de Versailles) : We consider the problem of estimatin
 g the Lévy density of a pure jump Lévy process\, possibly of infinite va
 riation\, from the high frequency observation of one trajectory. To direct
 ly construct an estimator of the Lévy density\, we use a compound Poisson
  approximation and we build a linear wavelet estimator. Its performance is
  studied in terms of $L_p$ loss functions\, $p\\geq1$\, over Besov balls. 
 To show that the resulting rates are minimax-optimal for a large class of 
 Lévy processes\, we propose new non-asymptotic bounds of the cumulative d
 istribution function of Lévy processes with Lévy density bounded from ab
 ove by the density of an alpha-stable type Lévy process in a neighbourhoo
 d of the origin. It is a joint work with Céline Duval.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle Döblin\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle Döblin:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2026@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20221208T104500
DTEND;TZID=Europe/Paris:20221208T114500
DTSTAMP:20221119T191842Z
URL:https://iecl.univ-lorraine.fr/events/lionel-cucala/
SUMMARY:Détection d'agrégats spatiaux : des statistiques de balayage pour
  données multivariées et fonctionnelles
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Lionel 
 Cucala (Université de Montpellier) : Dans ce travail\, nous nous intéres
 sons à des observations associées à une localisation spatiale (généra
 lement une position géographique) et nous cherchons à identifier des agr
 égats spatiaux\, i.e. des zones où les observations ont un comportement 
 atypique. Pour cela\, nous utilisons des méthodes de balayage spatial.\nA
 près avoir expliqué comment ces méthodes fonctionnent lorsque les obser
 vations sont réelles\, nous introduisons des statistiques conçues spéci
 fiquement pour le cas multivarié\, puis pour le cas fonctionnel.\nCes mé
 thodes sont appliquées sur des jeux de données environnementaux (concent
 ration de métaux polluants) et socio-économiques (taux de chômage).
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2001@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20221215T104500
DTEND;TZID=Europe/Paris:20221215T114500
DTSTAMP:20221208T141735Z
URL:https://iecl.univ-lorraine.fr/events/anna-melnykova/
SUMMARY:Non-asymptotic statistical test for the diffusion coefficient in st
 ochastic differential equations
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Anna Me
 lnykova (Université d'Avignon) : We develop a statistical test on the det
 erminant of the diffusion coefficient  in a 1 or 2-dimensional stochastic 
 differential equations from discrete observations on a fixed time interval
  [0\,T] sampled with a fixed time step.\nWe propose a test statistic based
  on increments of the process which guarantees the control of the level of
  the test in a non asymptotic setting. In  dimension 1\, the  test density
  is known explicitly even when the drift is  estimated. We construct the t
 est and give conditions under which the Type I and Type II errors can be c
 ontrolled. In dimension 2\, the test statistic has not an explicit density
  but upper and lower bounds are provided. We then give conditions under wh
 ich the Type I and Type II errors of the test procedure can be controlled.
  A numerical study illustrates the properties of the tests for  stochastic
  processes with known or unknown drifts.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2046@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20230112T104500
DTEND;TZID=Europe/Paris:20230112T114500
DTSTAMP:20230102T221324Z
URL:https://iecl.univ-lorraine.fr/events/vincent-bansaye/
SUMMARY:Processus de branchement pour des structures de contacts en épidé
 miologie
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Vincent
  Bansaye (École polytechnique) : Les processus de branchement apparaissen
 t dans de nombreux modèles de dynamique des populations\, notamment pour 
 décrire des invasions. En particulier\, ils interviennent dans la descrip
 tion des premières phases d'une épidémie pour déterminer si le nombre 
 d’infectés va exploser ou non et si oui à quelle vitesse. Dans ces mod
 èles\, la description des contacts joue un rôle important.\nAprès une i
 ntroduction sur ces problématiques\, nous nous focaliserons sur un modèl
 e incluant le traçage des contacts. Dans ce modèle\, les individus infec
 tent en population mélangée à taux fixe et l’information sur le conta
 ct infectieux est perdu à taux fixe\, tandis que le test d’un individu 
 infecté aboutit à l’isolement de la composante connexe associée aux c
 ontacts encore connus. Grâce à une propriété d’«éclatement» des a
 rbres récursifs uniformes\, nous pourrons réduire le modèle à un proce
 ssus de croissance fragmentation isolation sur les tailles des composantes
 .\nNous exploiterions alors des techniques récentes d’analyse quantitat
 ive des semi groupes non conservatifs et des processus de branchement asso
 cié. Cela permettra d’obtenir des convergences fortes pour décrire la 
 propagation de l’épidémie.\nEnfin\, nous évoquerons des extensions de
  ce modèle et la prise en compte d’une structuration spatiale des conta
 cts via de grands graphes aléatoires spatialisés\, impliquant des techni
 ques d’homogénéisation stochastique. \nCes travaux sont respectiveme
 nt des collaborations avec Chenlin Gu (Pékin) et Linglong Yuan (Liverpool
 )\, Michele Salvi (Rome) et Elisabeta Vergu (INRAe Jouy).
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2129@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20230126T104500
DTEND;TZID=Europe/Paris:20230126T114500
DTSTAMP:20230118T173744Z
URL:https://iecl.univ-lorraine.fr/events/henri-elad-altman/
SUMMARY:Temps locaux de processus et limites globales de leurs fonctionnell
 es additives
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Henri E
 lad Altman (Freie Universität Berlin) : Nous présenterons des résultats
  de limites d'échelles de fonctionnelles additives de processus non-marko
 viens\, dont nous décrirons la limite en termes de temps locaux du proces
 sus. Une étape clé dans nos preuves consiste en une nouvelle description
 s de fonctionnelles additives à l'aide d'un Lemme de la Couturière Stoch
 astique. Il s'agit de travaux en collaboration avec Khoa Lê (Université 
 de Leeds).
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle Döblin\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle Döblin:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2003@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20230302T104500
DTEND;TZID=Europe/Paris:20230302T114500
DTSTAMP:20230223T185520Z
URL:https://iecl.univ-lorraine.fr/events/alice-contat/
SUMMARY:Retrouver Adam et Eve dans les arbres de Barabàsi—Albert
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Alice C
 ontat (Université Paris-Saclay) : On se donne $(T (n) : n \\geq 1)$ un pr
 ocessus d’arbres construits récursivement sommet par sommet (par exempl
 e\, le processus de Barabàsi—Albert)\, que l’on observe à un temps n
  long. Notre but est de retrouver le sommet initial (Adam). Plus précisé
 ment\, on veut trouver un sous ensemble de sommets le plus petit possible 
 qui contient Adam avec probabilité au moins $1- \\varepsilon$.\nAprès un
  aperçu des résultats existants pour les arbres récursifs uniformes et 
 ceux de Barabàsi—Albert\, je montrerai que pour trouver Adam\, il vaut 
 mieux chercher Adam et Eve.\nTravail en collaboration avec Nicolas Curien\
 , Perrine Lacroix\, Etienne Lasalle et Vincent Rivoirard.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2047@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20230323T104500
DTEND;TZID=Europe/Paris:20230323T114500
DTSTAMP:20230320T225202Z
URL:https://iecl.univ-lorraine.fr/events/nabil-kazi-tani/
SUMMARY:Le rôle de la corrélation dans les jeux différentiels stochastiq
 ues
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Nabil K
 azi-Tani (IECL) : Dans cet exposé\, on s'intéressera aux équilibres de 
 Nash dans des jeux différentiels à deux joueurs avec contrôle sur les c
 oefficients de diffusion\, et avec corrélation des mouvements browniens d
 irigeant les équations. Nous considérerons des jeux de classement à som
 me nulle\, dans le sens où les critères optimisés par les joueurs ne d
 épendent que de la différence des processus d'état des deux joueurs. Da
 ns ce cadre\, nous donnerons explicitement les stratégies d'équilibres\,
  en fonction de la corrélation : si le coefficient de corrélation est in
 férieur à un certain seuil\, les stratégies d'équilibres sont des "con
 trôles forts"\, tandis que si le coefficient de corrélation est supérie
 ur à ce même seuil\, les stratégies d'équilibres sont "mixtes"\, dans 
 le sens où les joueurs auront un intérêt à choisir leurs stratégies a
 u hasard. Cet exposé sera l'occasion d'introduire la formulation dite "re
 laxée" des problèmes de contrôle et de jeu stochastiques\, formulation 
 basée sur des solutions à des problèmes de martingales. Il s'agit d'un 
 travail en collaboration avec Stefan Ankirchner (Université de Jena) et J
 ulian Wendt (Université de Jena).
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2007@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20230330T104500
DTEND;TZID=Europe/Paris:20230330T114500
DTSTAMP:20230323T121222Z
URL:https://iecl.univ-lorraine.fr/events/mouad-ramil-2/
SUMMARY:Estimées bilatérales pour le processus de Langevin tué au bord d
 'un domaine
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Mouad R
 amil (Seoul National University) : Nous nous intéresserons ici à l'obten
 tion d'estimées bilatérales pour la densité du processus de Langevin tu
 é au bord de l'ouvert D=(0\,1) x R^d dans le cas d'un potentiel quadratiq
 ue. Ces estimées sont cruciales pour obtenir un résultat de convergence 
 de la distribution conditionnée à rester dans le domaine D vers l'unique
  distribution quasi-stationnaire\, avec un préfacteur indépendent de la 
 distribution initiale. En raison de la dégénérescence du Langevin et du
  fait que le domaine D soit non borné\, il est nécessaire de développer
  une approche nouvelle. Dans ce travail\, nous commençons par prouver des
  estimées bilatérales et explicites sur la probabilité du premier temps
  de sortie\, que nous parvenons ensuite à étendre à la densité du proc
 essus tué à l'aide de contrôles gaussiens obtenus dans un travail préc
 édent.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2086@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20230406T104500
DTEND;TZID=Europe/Paris:20230406T114500
DTSTAMP:20230330T123658Z
URL:https://iecl.univ-lorraine.fr/events/lucile-laulin/
SUMMARY:La limite super-diffusive de la marche aléatoire de l'éléphant
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Lucile 
 Laulin (Université de Nantes) : La marche aléatoire de l'éléphant (ERW
 ) est une marche aléatoire discrète qui a été introduite au début des
  années 2000 par deux physiciens afin d'étudier l’influence d’un par
 amètre de mémoire sur le comportement de la marche aléatoire. On commen
 cera par introduire la marche aléatoire de l’éléphant en dimension 1 
 et sa généralisation à dimension d. On présentera ensuite plusieurs po
 ssibilités pour étudier et obtenir des résultats sur l’ERW\, telles q
 ue l’approche martingale\, lien avec les urnes de Polya ou encore les ar
 bres aléatoires récursifs. En particulier\, on expliquera comment l’ut
 ilisation des trois approches est nécessaire pour obtenir des information
 s sur la variable aléatoire limite qui apparaît dans le régime super-di
 ffusif en dimension 1\, et éventuellement en dimension d.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2027@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20230413T104500
DTEND;TZID=Europe/Paris:20230413T114500
DTSTAMP:20230408T212835Z
URL:https://iecl.univ-lorraine.fr/events/maximilian-engel/
SUMMARY:An ergodic theory for conditioned random dynamical systems
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Maximil
 ian Engel (Freie Universität Berlin) : We will introduce the setting of r
 andom dynamical systems with absorption\, as for example induced by iterat
 ed function systems or stochastic differential equations on a bounded doma
 in with killing at the boundary. We will show how to embed the process con
 ditioned to never being absorbed\, the Q-process\, into the framework of r
 andom dynamical systems\, allowing us to study multiplicative ergodic prop
 erties and the notion of Kolmogorov-Sinai entropy. Finally\, we will elabo
 rate on the relation between entropy and Lyapunov exponents for this setti
 ng.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2124@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20230504T104500
DTEND;TZID=Europe/Paris:20230504T114500
DTSTAMP:20230425T152731Z
URL:https://iecl.univ-lorraine.fr/events/mahendra-mariadassou/
SUMMARY:The Poisson-lognormal model as a versatile framework for the joint 
 analysis of species abundances
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Mahendr
 a Mariadassou (Inrae Jouy-en-Josas) : Joint work with Julien Chiquet and S
 téphane Robin\n\nJoint Species Abundance Models (JSDM) provide a general 
 multivariate framework to study the joint abundances of all species from a
  community. JSDM account for both structuring factors (environmental chara
 cteristics or gradients\, such as habitat type or nutrient availability) a
 nd potential interactions between the species (competition\, mutualism\, p
 arasitism\, etc.)\, which is instrumental in disentangling meaningful ecol
 ogical interactions from mere statistical associations.\n\nModeling the de
 pendency between the species is challenging because of the count-valued na
 ture of abundance data and most JSDM rely on Gaussian latent layer to enco
 de the dependencies between species in a covariance matrix. The multivaria
 te Poisson-lognormal (PLN) model is one such model\, which can be viewed a
 s a multivariate mixed Poisson regression model. The inference of such mod
 els raises both statistical and computational issues\, many of which were 
 solved in recent contributions using variational techniques and convex opt
 imization.\n\nThe PLN model turns out to be a versatile framework\, within
  which a variety of analyses can be performed\, including multivariate sam
 ple comparison\, clustering of sites or samples\, dimension reduction (ord
 ination) for visualization purposes\, or inference of interaction networks
 . We will presents the PLN framework and some variants and illustrate them
  on typical datasets. All the models and methods are implemented in the R 
 package PLNmodels\, available from cran.r-project.org.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2111@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20230511T104500
DTEND;TZID=Europe/Paris:20230511T114500
DTSTAMP:20230505T102802Z
URL:https://iecl.univ-lorraine.fr/events/joseph-najnudel/
SUMMARY:Convergence de fonctions holomorphes aléatoires
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Joseph 
 Najnudel (Université de Nice Sophia-Antipolis) : Dans cet exposé\, nous 
 étudions les conditions sous lesquelles la convergence du processus ponct
 uel des zéros de fonctions holomorphes aléatoire implique la convergence
  des fonctions holomorphes elles-mêmes. Nous montrons comment appliquer c
 e résultat au cas du polynome caractéristique de matrices aléatoires\, 
 et conjecturalement\, à la fonction zeta de Riemann.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle Döblin\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle Döblin:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2106@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20230525T104500
DTEND;TZID=Europe/Paris:20230525T114500
DTSTAMP:20230517T184025Z
URL:https://iecl.univ-lorraine.fr/events/benoit-corsini/
SUMMARY:Modèles aléatoires d'arbres binaires de recherche
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Benoît
  Corsini (Eindhoven University of Technology) : Lors de ce séminaire je v
 ais présenter différents modèles d'arbres binaires de recherche\, un ob
 jet couramment utilisé en informatique pour étudier l'organisation optim
 ale de données. Je vais commencer par définir plusieurs modèles d'arbre
 s binaires de recherche\, basés sur différents modèles de permutations 
 (permutations uniformes\, de Mallows et biaisées de record). Après ça\,
  je vais expliquer quelques propriétés intéressantes de ces arbres\, no
 tamment comment ils peuvent être construits récursivement. Finalement\, 
 je vais conclure cette présentation en donnant des résultats concernant 
 la hauteur de ces arbres et expliquer comment ces résultats peuvent être
  extraits des constructions précédemment introduites.\n\nL'exposé s'int
 ègrera à la masterclass M2
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2160@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20230601T104500
DTEND;TZID=Europe/Paris:20230601T114500
DTSTAMP:20230524T110223Z
URL:https://iecl.univ-lorraine.fr/events/laure-mareche/
SUMMARY:Théorèmes limites pour une marche aléatoire avec auto-interactio
 ns
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Laure M
 arêché (Université de Strasbourg) : Dans cet exposé\, on s'intéresser
 a à une marche aléatoire non markovienne\, telle que la probabilité que
  la marche aille en un point donné est plus faible si elle est déjà pas
 sée souvent le long de l'arête entre la position initiale et la cible. L
 es modèles de ce type les plus étudiés sont ceux dans lesquels les arê
 tes sont non orientées. Cependant\, en 2008 Tóth et Vető ont introduit 
 une telle marche aléatoire avec auto-interactions pour laquelle les arêt
 es sont orientées\, et découvert des propriétés très différentes de 
 celles des modèles avec arêtes non orientées. Malgré l'intérêt d'un 
 tel comportement\, très peu de résultats ont été obtenus depuis sur ce
  modèle. On présentera de nouveaux théorèmes limites pour cette marche
  aléatoire.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle Döblin\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle Döblin:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2025@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20230608T104500
DTEND;TZID=Europe/Paris:20230608T114500
DTSTAMP:20230118T171945Z
URL:https://iecl.univ-lorraine.fr/events/barbara-pascal/
SUMMARY:The Kravchuk transform: a novel covariant representation for discre
 te signals amenable to zero-based detection tests
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Barbara
  Pascal (LS2N\, Nantes) : Recent works in time-frequency analysis proposed
  to switch the focus from the maxima of the spectrogram toward its zeros\,
  which form a random point pattern with a very stable structure. Several s
 ignal processing tasks\, such as component disentanglement and signal dete
 ction procedures\, have already been renewed by using modern spatial stati
 stics onthe pattern of zeros. Tough\, they require cautious choice of both
  the discretization strategy and the observation window in the time-freque
 ncy plane. To overcome these limitations\, we propose a generalized time-f
 requency representation: the Kravchuk transform\, especially designed for 
 discrete signals analysis\, whose phase space is the unit sphere\, particu
 larly amenable to spatial statistics. We show that it has all desired prop
 erties for signal processing\, among which covariance\, invertibility and 
 symmetry\, and that the point process of the zeros of the Kravchuk transfo
 rm of complex white Gaussian noise coincides with the zeros of the spheric
 al Gaussian Analytic Function. Elaborating on this theorem\, we finally de
 velop a Monte Carlo envelope test procedure for signal detection based on 
 the spatial statistics of the zeros of the Kravchuk spectrogram.\n\nAfter 
 reviewing the unorthodox path focusing on the zeros of the standard spectr
 ogram and the associated theoretical results on the distribution of zeros 
 in the case of white noise\, I will introduce the Kravchuk transform and s
 tudy the random point process of its zeros from a spatial statistics persp
 ective. Then I will present the designed Monte Carlo envelop test\, and il
 lustrate its numerical performance in adversarial settings\, with both low
  signal-to-noise ratio and small number of samples\, and compare it to sta
 te-of-the-art zeros-based detection procedures.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2289@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20230927T104500
DTEND;TZID=Europe/Paris:20230927T114500
DTSTAMP:20230915T142719Z
URL:https://iecl.univ-lorraine.fr/events/jochen-blath/
SUMMARY:Probabilistic structures emerging from dormancy
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Jochen 
 Blath (Frankfurt) : Dormancy is a complex trait that has evolved independe
 ntly many times across the tree of life. In particular many micro-organism
 s can enter a reversible state of vanishing metabolic activity. The corres
 ponding dormancy periods can range from a few hours to potentially thousan
 ds of years. Also\, the dormancy transitioning mechanisms are highly diver
 se\, including spontaneous dormancy initiation and resuscitation\, respons
 ive switching due to environmental cues\, and competition-induced dormancy
  initiation. In general\, dormancy allows a population to maintain a reser
 voir of genotypic and phenotypic diversity (that is\, a seed bank) that ca
 n contribute to its longterm survival and coexistence. In this talk\, we r
 eview some of the probabilistic structures that emerge from stochastic ind
 ividual based models involving dormancy.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2253@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20230928T104500
DTEND;TZID=Europe/Paris:20230928T114500
DTSTAMP:20230915T202053Z
URL:https://iecl.univ-lorraine.fr/events/masato-hoshino/
SUMMARY:BPHZ theorem on regularity-integrability structures
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Masato 
 Hoshino (Osaka University) : In the study of singular SPDEs\, it has been 
 a challenging problem to obtain a simple proof of a general probabilistic 
 convergence result (BPHZ theorem). Differently from Chandra and Hairer's F
 eynman diagram approach\, Linares\, Otto\, Tempelmayr\, and Tsatsoulis rec
 ently proposed an inductive proof based on the spectral gap inequality by 
 using their multiindex language. Inspired by their approach\, Hairer and S
 teele also obtained an inductive proof by using the regularity structure l
 anguage. In this talk\, we introduce an extension of the regularity struct
 ure including integrability exponents and provide a simpler proof of BPHZ 
 theorem.\nThis talk is based on a joint work with Ismael Bailleul (Univ Br
 est).
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2279@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20231005T104500
DTEND;TZID=Europe/Paris:20231005T114500
DTSTAMP:20230919T094459Z
URL:https://iecl.univ-lorraine.fr/events/valentin-rapenne/
SUMMARY:Processus renforcés\, champs aléatoires et limites d'échelle
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Valenti
 n Rapenne (IECL) : Le Processus de saut renforcé par sommet (VRJP en rais
 on de son acronyme anglais) est un processus aléatoire en temps continu a
 uto-renforcé défini sur un graphe : plus il passe par un sommet\, plus i
 l est probable qu'il y revienne à l'avenir. Depuis les travaux de Sabot e
 t Tarrès\, on sait que le VRJP peut être vu comme un processus Markovien
  en environnement aléatoire. Cet environnement est caractérisé par un c
 ertain champ aléatoire U qui\, étonnamment\, était déjà connu par les
  physiciens en tant que modèle sigma super-symétrique H^{(2|2)}\, un obj
 et important en physique quantique. Par un changement de variable\, on peu
 t ramener l'étude du champ U à celle d'un autre champ beta. Ce champ bet
 a permet de définir un opérateur de Schrödinger aléatoire H dont les p
 ropriétés sont reliées au VRJP. H est notamment inversible et son inver
 se G permet de retrouver l'environnement du VRJP. Dans cet exposé\, nous 
 expliquerons ce cheminement allant du VRJP vers les champs aléatoires et 
 nous expliquerons les résultats connus sur ces champs et sur le VRJP. Pou
 r finir\, nous parlerons d'un résultat plus récent établissant la limit
 e d'échelle de G sur le cercle. Cela permet de définir un opérateur de 
 Schrödinger aléatoire continu analogue à H sur le cercle.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2254@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20231012T104500
DTEND;TZID=Europe/Paris:20231012T114500
DTSTAMP:20230929T102949Z
URL:https://iecl.univ-lorraine.fr/events/tara-trauthwein/
SUMMARY:Normal Approximation of Poisson Functionals via Malliavin-Stein Met
 hod
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Tara Tr
 authwein (Université du Luxembourg) : Imagine a collection of randomly di
 stributed points in space\, and build a graph on it according to some rule
 s. We now take the sum of all edge lengths and want to know if this sum ve
 rifies a central limit theorem. In our context\, the random collection of 
 points is given by a Poisson measure and the sum of edge lengths is an exa
 mple of a Poisson functional. In this talk\, using the so-called Malliavin
 -Stein method\, we show a result which allows us to derive central limit t
 heorems\, as well as speeds of convergence\, for functionals of general Po
 isson measures\, under minimal moment assumptions. Our main application is
  the closure of a conjecture about a convergence result of the Online Near
 est Neighbour Graph.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2276@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20231019T104500
DTEND;TZID=Europe/Paris:20231019T114500
DTSTAMP:20231004T183723Z
URL:https://iecl.univ-lorraine.fr/events/celine-comte/
SUMMARY:Stochastic dynamic matching – A mixed graph-theory and linear-alg
 ebra approach
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Céline
  Comte (LAAS-CNRS Toulouse) : The stochastic dynamic matching problem has 
 recently drawn attention in the stochastic-modeling community due to its n
 umerous applications\, ranging from supply-chain management to kidney exch
 ange programs. In this presentation\, we consider a matching problem in wh
 ich items of different classes arrive according to independent Poisson pro
 cesses\, unmatched items are stored in a queue\, and compatibility constra
 ints are described by a simple graph on the classes\, so that two items ca
 n be matched if their classes are neighbors in the graph. We analyze the e
 fficiency of matching policies\, not only in terms of system stability\, b
 ut also in terms of matching rates between different classes. Our results 
 rely on the observation that\, under any stable policy\, the matching rate
 s satisfy a conservation equation that equates the arrival and departure r
 ates of each item class.\n\nThis presentation is based on a joint work wit
 h Fabien Mathieu (LINCS) and Ana Bušić (Inria and PSL University). A pre
 print is available at the following address: https://arxiv.org/abs/2112.14
 457.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2365@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20231109T104500
DTEND;TZID=Europe/Paris:20231109T114500
DTSTAMP:20231106T123243Z
URL:https://iecl.univ-lorraine.fr/events/un-phenomene-de-concentration-en-
 geometrie-combinatoire/
SUMMARY:Un phénomène de concentration en géométrie combinatoire
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Xavier 
 Goaoc (Gamble\, LORIA) : \n\nLe type d’ordre d’une séquence de points
  du plan est une généralisation de la permutation associée à une séqu
 ence de nombres réels. Cette objet combinatoire encode de nombreuses prop
 riétés géométriques de la séquence de points\, par exemple le treilli
 s des faces de son enveloppe convexe\, ou encore les triangulations qu’e
 lle supporte. Cet exposé commencera par une rapide introduction à ces ob
 jets. Je discuterai ensuite d’un phénomène de concentration qui appara
 ît lorsque l’on lit les types d’ordres de séquences de points aléat
 oires\, pour divers modèles naturels. Cette concentration rend difficile 
 une bonne exploration aléatoire de ces structures.\n\nCeci est un travail
  conjoint avec Emo Welzl\n\nhttps://dl.acm.org/doi/10.1145/3570636\nhttps:
 //arxiv.org/abs/2003.08456\n\n\n
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2264@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20231116T104500
DTEND;TZID=Europe/Paris:20231116T114500
DTSTAMP:20231107T214701Z
URL:https://iecl.univ-lorraine.fr/events/davide-giraudo/
SUMMARY:Théorème limite central fonctionnel et loi des grands nombres pou
 r des U-statistiques à valeur dans un espace de Hilbert
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Davide 
 Giraudo (Université de Strasbourg) : Après avoir introduit les U-statist
 iques de données indépendantes\, le théorème limite central fonctionne
 l et la loi des grands nombres correspondantes\, nous présenterons des r
 ésultats récents pour des U-statistiques basées sur des suites stationn
 aires dépendantes à valeurs dans un espace métrique séparable. Nous tr
 aiterons le cas des suites beta-mélangeantes (qui se prêtent bien au cou
 plage) ainsi que celui des U-statistiques à valeurs dans un espace de Hil
 bert. Ces dernières se révèlent utiles dans certains tests statistiques
 .
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2266@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20231123T104500
DTEND;TZID=Europe/Paris:20231123T114500
DTSTAMP:20231113T185736Z
URL:https://iecl.univ-lorraine.fr/events/felix-cheysson/
SUMMARY:Spectral estimation for Hawkes processes
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Felix C
 heysson (Université Gustave Eiffel) : Hawkes processes are a family of po
 int processes for which the occurrence of any event increases the probabil
 ity of further events occurring. Although the linear Hawkes process\, for 
 which a representation in the form of a superposition of branching process
 es exists\, is particularly well studied\, difficulties remain in estimati
 ng the parameters of the process from imperfect data (noisy\, missing or a
 ggregated data)\, since the usual estimation methods based on maximum like
 lihood or least squares do not necessarily offer theoretical guarantees or
  are numerically too costly.\nIn this work\, we propose a spectral approac
 h well-adapted to this context\, for which we prove consistency and asympt
 otic normality. In order to derive these properties\, we show that Hawkes 
 processes can be studied through the scope of mixing\, opening the use of 
 central limit theorems that already exist in the literature.\nI will then 
 present two applications of this approach: to aggregated data (joint work 
 with Gabriel Lang)\; and to noisy data (joint work with Anna Bonnet\, Migu
 el Martinez and Maxime Sangnier).
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2374@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20231130T091500
DTEND;TZID=Europe/Paris:20231130T101500
DTSTAMP:20231128T135928Z
URL:https://iecl.univ-lorraine.fr/events/frederik-ravn-klausen/
SUMMARY:Phase transitions of the graphical representations of the Ising mod
 el
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Frederi
 k Ravn Klausen (University of Copenhagen) : After much success in using th
 e double random current representation in the study of the Ising model\, D
 uminil-Copin posed the question in 2016 of determining the (percolative) p
 hase transition of the single random current. By relating the single rando
 m current to the loop O(1) model (also known as the high-temperature expan
 sion and Eulerian percolation)\, we prove polynomial lower bounds for path
  probabilities (and infinite expectation of cluster sizes) for both the si
 ngle random current and loop O(1) model corresponding to any supercritical
  Ising model on the hypercubic lattice. Thereby partially resolving the p
 osed question.\n\nIn this talk\, I will gently introduce graphical represe
 ntations of the Ising model\, their monotonicity properties and relations 
 through Bernoulli sprinkling and the uniform even subgraph. Afterward\, we
  discuss new results whose surprising proof takes inspiration from the tor
 ic code in quantum theory. Based on joint work with Ulrik Tinggaard Hansen
  and Boris Kjær: https://arxiv.org/abs/2306.05130.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2287@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20231130T104500
DTEND;TZID=Europe/Paris:20231130T114500
DTSTAMP:20231124T090615Z
URL:https://iecl.univ-lorraine.fr/events/la-construction-et-la-validation-
 dun-lexique-didees-de-la-politique-daristote/
SUMMARY:Quelle est la probabilité que les Spartiates aient été de bons c
 uisiniers ?
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Pierre 
 Mercuriali (IECL) : Nous cherchons à comprendre et à automatiser la mani
 ère dont les historiens comprennent un texte historique. Partons du princ
 ipe que lors de la lecture du texte\, un historien manipule un ensemble d'
 hypothèses desquelles découlent autant de mondes possibles : si Caton es
 t né à une certaine date\, alors son parcours académique est exemplaire
  et correspond aux standards de l'époque \; sinon\, son parcours est sign
 ificatif en ce sens qu'il est unique parmi ses pairs.\nJe montrerai en quo
 i les logiques modales\, auxquelles on attribue une sémantique probabilis
 te\, peuvent modéliser un ensemble d'hypothèses de nature historique. J'
 illustrerai par un cas test sur la Politique d'Aristote et\, en particulie
 r\, la notion de «συσσιτία»\, ou de « réunion de convives ».\
 n
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2280@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20231207T104500
DTEND;TZID=Europe/Paris:20231207T114500
DTSTAMP:20231127T082223Z
URL:https://iecl.univ-lorraine.fr/events/xue-mei-li/
SUMMARY:Correlated noises in stochastic differential equations
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Xue-Mei
  Li (Imperial College London & EPFL) : It is a standard assumption that th
 e Gaussian noises in stochastic systems are white in time and white space.
  This means that the noise at different point in space or in time are assu
 med to be uncorrelated. This leads to the Ito theory of integration. Howev
 er\, some time series data and other data indicate otherwise\, some even e
 xhibits long range dependence. In SDEs these imply that neither the Markov
  theory nor its martingale characterisation can be relied on. In SPDEs\, t
 he difficulty of irregularity coming from the white noise can be mitigated
  if they are replaced by smooth correlated noise. But other problems arriv
 e. In this talk we shall explore these models and some phenomenons. New\, 
 as well as old\, techniques in Stochastic Analysis will be explored.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2358@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20231214T104500
DTEND;TZID=Europe/Paris:20231214T114500
DTSTAMP:20231201T134415Z
URL:https://iecl.univ-lorraine.fr/events/meltem-unel/
SUMMARY:La limite locale des arbres pondérés exponentiellement par la hau
 teur
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Meltem 
 Unel (Orsay) : Le cas le plus simple et peut-être le plus naturel des lim
 ites locales des arbres est Uniform Infinite Planar Tree: on commence par 
 la suite des mesures de probabilité uniforme \nu_N dont le support est l
 ’ensemble des arbres plans enracinés de taille N et on étudie la limit
 e faible \nu de cette suite\, dont le support est l’ensemble des arbres 
 plans enracinés de taille infinie.\nUne modification naturelle dans la re
 cherche des limites différentes est de pondérer les arbres : est-ce que 
 la nouvelle suite des mesures \\rho_N\, par rapport à laquelle la valuer 
 d’un arbre de taille N est proportionnelle à son poids\, admet une limi
 te faible ?\nDans cet exposé\, on considère des arbres planes enracin
 és dont la distribution est uniforme pour une hauteur h et une taille N f
 ixée et dont la dépendance à la hauteur est de forme exponentielle\, \\
 exp(-\\mu h)\, pour \\mu réel. En définissant le poids total de ces arbr
 es de taille N fixe comme Z^{\\mu}_N\, on détermine son comportement asym
 ptotique pour N grand\, pour \\mu réel quelconque. Finalement\, on identi
 fie la limite locale des mesures de probabilité correspondantes et trouve
  une transition à \\mu=0 d'une phase à une seule épine à une phase à 
 plusieurs épines (backbone). En conséquence\, il y a une transition dans
  le taux de croissance du volume des boules autour de la racine en fonctio
 n du rayon\, passant d'une croissance linéaire pour \\mu &lt\; 0 à la cr
 oissance quadratique familière à \\mu=0 et à une croissance cubique pou
 r \\mu &gt\; 0.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2425@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20240111T091500
DTEND;TZID=Europe/Paris:20240111T101500
DTSTAMP:20240109T163753Z
URL:https://iecl.univ-lorraine.fr/events/limite-locale-des-animaux-diriges
 -dans-le-quart-de-plan/
SUMMARY:Limite locale des animaux dirigés dans le quart de plan
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Arvind 
 Singh : On appelle "animal dirigé" un sous-ensemble fini du quart de plan
  N x N qui contient l'origine et tel que tout autre site possède au moins
  un voisin à sa gauche ou en dessous de lui. Dans cet exposé\, je regard
 erai la limite quand n tend vers l'infini d'un animal choisi uniformément
  parmi les animaux à n sommets. Je montrerai en particulier que l'objet l
 imite est encodé par une marche aléatoire et peut aussi s’interpréter
  comme un système de particules en interaction possédant une remarquable
  propriété de Markov spatiale (travail en collaboration avec O. Hénard 
 et E. Maurel-Segala).
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2265@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20240111T104500
DTEND;TZID=Europe/Paris:20240111T114500
DTSTAMP:20240104T095717Z
URL:https://iecl.univ-lorraine.fr/events/ivan-nourdin/
SUMMARY:TCL quantitatifs pour réseaux neuronaux profonds
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Ivan No
 urdin (Université du Luxembourg) : Dans cet exposé\, nous étudierons le
  comportement asymptotique des réseaux neuronaux entièrement connectés 
 avec poids et biais gaussiens et dont la taille des couches cachées tend 
 vers l'infini. Basé sur un travail commun récent avec S. Favaro\, B. Han
 in\, D. Marinucci et G. Peccati.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2272@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20240118T104500
DTEND;TZID=Europe/Paris:20240118T114500
DTSTAMP:20240110T094748Z
URL:https://iecl.univ-lorraine.fr/events/manon-costa-2/
SUMMARY:Cycles éco-évolutifs dans des communautés proies-prédateurs
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Manon C
 osta (Université Paul Sabatier) : Dans cet exposé\, nous présentons et 
 étudions un modèle pour deux populations avec une interaction prédateur
 -proie\, où chaque population est composée de deux types d'individus\, n
 otés 0 et 1\, de sorte que les prédateurs d'un type donné prospèrent e
 n présence de proies similaires\, tandis que les proies d'un type donné 
 ont plus de chances de survivre en présence de prédateurs du type diffé
 rent.\nNous considérons une limite dans une grande population avec des mu
 tations dans à une échelle intermédiaire\, c’est à dire que le taux 
 de mutation individuel disparaît tandis que le taux de mutation total ten
 d vers l'infini. Nous prouvons qu'en fonction des paramètres du modèle\,
  différents scénarios peuvent se produire : invasion successive de proie
 s et de prédateurs conduisant à la coexistence de quatre types\, ou inva
 sion successive de proies dans une population de prédateurs résidents co
 nduisant soit à l'extinction des proies\, soit à la coexistence de tous 
 les types\, ...
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle Döblin\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle Döblin:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2328@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20240125T104500
DTEND;TZID=Europe/Paris:20240125T114500
DTSTAMP:20240123T101618Z
URL:https://iecl.univ-lorraine.fr/events/carlo-bellingeri/
SUMMARY:Formule d'Euler-Maclaurin et intégrales itérées généralisées
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Carlo B
 ellingeri : Considérée comme l'une des identités clés de l'analyse cla
 ssique\, la formule d'Euler-McLaurin est l'un des outils standard pour rel
 ier les sommes et les intégrales\, avec des applications remarquables dan
 s de nombreux domaines des mathématiques\, bien que peu utilisée en anal
 yse stochastique. Dans cet exposé\, nous montrerons comment\, en introdui
 sant de nouvelles variantes des intégrales itérées d'un chemin et un si
 mple problème variationnel\, nous pouvons généraliser cette identité d
 ans le contexte de l'intégration de Riemann Stieltjes.\n\n(Le séminaire 
 aura lieu en amphi 3.)
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:2318@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20240201T104500
DTEND;TZID=Europe/Paris:20240201T114500
DTSTAMP:20240122T075734Z
URL:https://iecl.univ-lorraine.fr/events/lisa-hartung/
SUMMARY:Maxima of a random model of the Riemann zeta function on longer int
 ervals (and branching random walks)
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Lisa Ha
 rtung : We study the maximum of a random model for the Riemann zeta functi
 on (on the critical line at height T) on the interval $[-(\\log T)^\\theta
 \,(\\log T)^\\theta]$\, where $ \\theta= = (\\log \\log T)^{-a}$\, with $0
 &lt\;a&lt\;1$.  We obtain the leading order as well as the logarithmic co
 rrection of the maximum.\nAs it turns out\, a good toy model is a collecti
 on of independent BRWs\, where the number of independent copies depends on
  $\\theta$. In this talk I will try to motivate our results by mainly focu
 sing on this toy model. The talk is based on joint work in progress with L
 .-P. Arguin and G. Dubach.\n(Séminaire commun avec l'équipe ATN.)
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2330@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20240208T104500
DTEND;TZID=Europe/Paris:20240208T114500
DTSTAMP:20240126T080558Z
URL:https://iecl.univ-lorraine.fr/events/sandro-franceschi/
SUMMARY:Mouvement brownien réfléchi dans un cône : étude du cas transie
 nt
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Sandro 
 Franceschi : La littérature consacrée au mouvement brownien réfléchi d
 ans un cône bidimensionnel est la plupart du temps consacrée à l'étude
  de sa distribution stationnaire dans le cas récurrent. Dans cet exposé\
 , nous intéresserons en revanche au cas transient pour étudier les fonct
 ions de Green de ce processus et leurs asymptotiques. Ceci nous amènera 
 à considérer la frontière de Martin associée et les fonctions harmoniq
 ues satisfaisant des conditions de Neumann obliques sur les bords du cône
 . Pour certains modèles\, nous illustrerons cela en étudiant la probabil
 ité d'évasion du processus le long d'un axe et sa probabilité d'absorpt
 ion au sommet du cône.\n\nPour établir nos résultats\, nous utilisons d
 es méthodes analytiques historiquement développées pour étudier les ma
 rches aléatoires dans le quadrant. Nous établissons des équations fonct
 ionnelles satisfaites par les transformées de Laplace des fonctions de Gr
 een. Grâce à la théorie des problèmes frontières (de Riemann et Carle
 man)\, il est possible de déterminer des formules explicites pour ces tra
 nsformées de Laplace impliquant des fonctions hypergéométriques. La mé
 thode du point col et des lemmes de transfert taubériens permettent d'obt
 enir des résultats asymptotiques et d'établir la frontière de Martin.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2271@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20240215T104500
DTEND;TZID=Europe/Paris:20240215T114500
DTSTAMP:20240207T201553Z
URL:https://iecl.univ-lorraine.fr/events/mike-pereira/
SUMMARY:Gaussian random fields on Riemannian manifolds: Applications to Geo
 statistics
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Mike Pe
 reira (Université Paris Sciences & Lettres) : Many applications in spatia
 l and spatio-temporal statistics require data to be modeled by Gaussian pr
 ocesses on non-Euclidean domains\, or with non-stationary properties. Usin
 g such models generally comes at the price of a drastic increase in operat
 ional costs (computational and storage-wise)\, rendering them hard to appl
 y to large datasets. In this talk\, we propose a solution to this problem\
 , which relies on the definition of a class of random fields on Riemannian
  manifolds. These fields extend ongoing work that has been done to leverag
 e a characterization of the random fields classically used in Geostatistic
 s as solutions of stochastic partial differential equations. The discretiz
 ation of these generalized random fields\, undertaken using a finite eleme
 nt approach\, then provides an explicit characterization that is leveraged
  to solve the scalability problem. Indeed\, matrix-free algorithms\, in th
 e sense that they do not require to build and store any covariance (or pre
 cision) matrix\, are derived to tackle for instance the simulation of larg
 e Gaussian fields with given covariance properties\, even in the non-stati
 onary setting or on surfaces.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2331@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20240222T104500
DTEND;TZID=Europe/Paris:20240222T114500
DTSTAMP:20240212T164645Z
URL:https://iecl.univ-lorraine.fr/events/adrien-cotil/
SUMMARY:Comportement en temps long des équations de Cucker-Smale et infér
 ence de structure sociale
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Adrien 
 Cotil : La compréhension de l'auto-organisation d'un système\, c'est-à-
 dire sa capacité à faire émerger des comportements collectifs sans inte
 rvention extérieure\, est à la base du développement de nombreux domain
 es scientifiques\, aussi bien en physique\, en informatique\, en mathémat
 iques\, en biologie ou en sociologie. Au sein de ce domaine se trouve l'é
 tude des modèles de consensus\, permettant de décrire comment des agents
  s'échangent de l'information afin d'aboutir à une décision commune. Da
 ns cet exposé\, nous aborderons un modèle de consensus largement étudi
 é dans la littérature : le modèle de Cucker-Smale. Ce dernier décrit d
 es individus qui se déplacent dans l'espace et qui s'alignent les uns sur
  les autres. Il suppose que la force avec laquelle les individus s'alignen
 t entre eux dépend à la fois de la distance qui les sépare et d'un para
 mètres A(i\,j) qui décrit intrinsèquement comment un individu i s'align
 e sur un individu j. L'une des questions principales est la détermination
  de conditions qui assurent que les individus tendent tous à se déplacer
  dans la même direction à la même vitesse\, appelé phénomène de floc
 king dans ce contexte. En exploitant la dualité entre les équations de C
 ucker-Smale et les équations de Kolomogorov\, nous prouvons que le flocki
 ng est équivalent à la convergence en variation total d’un certain pro
 cessus de saut markovien inhomogène en temps.  Nous prouvons ensuite cet
 te convergence en utilisant des techniques de type Doeblin\, permettant de
  dériver de nouvelles conditions de flocking plus fines que celles connue
 s pour ce modèle. Enfin\, nous traiterons la question de l'apprentissage 
 du paramètre A(i\,j) à partir de données de déplacement d'animaux\, pe
 rmettant d'obtenir des informations sur la manière dont ceux-ci se compor
 tent socialement les uns avec les autres.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2452@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20240314T091500
DTEND;TZID=Europe/Paris:20240314T101500
DTSTAMP:20240312T080215Z
URL:https://iecl.univ-lorraine.fr/events/hernan-mardones/
SUMMARY:Systems of FBSDEs driven by Brownian Motion and Numerical Simulatio
 n of Fluid Dynamics
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Hernán
  A. Mardones González (Universidad de la Frontera\, Chile) : The systems 
 of forward-backward stochastic differential equations driven by Brownian m
 otion (FBSDEs for short) help us to model diffusion processes related to p
 henomena that involve environment perturbations. The drift coefficients co
 nstitute the descriptive part of a non-random ambient\, while the Wiener p
 rocesses permit us to describe the random perturbations involved into the 
 dynamics through the diffusion terms. The systems of FBSDEs motion are lin
 ked to the nonlinear partial differential equations (PDEs) through the Fey
 man-Kac formulae. Therefore\, the deterministic solutions can be obtained 
 by probabilistic representations involving the stochastic processes that s
 olve the FBSDEs.\n\nDuring this talk\, we deal with the numerical simulati
 on of systems of stochastic particles ruled by FBSDEs associated with nonl
 inear PDEs appearing in fluid dynamics. To make this\, we discretize local
 ly in time the stochastic equations\, and then we consider integration sch
 emes of Euler-Maruyama type\, together with the optimal quantization of th
 e involved Wiener increments as an alternative to the Monte-Carlo simulati
 on. Then we approximate the related conditional expectations over each tem
 poral-spatial node of a computational domain with uniform discretization s
 teps in time and space. Numerical results are presented to the case of ana
 lytic spatially-periodic exact solutions of the incompressible Navier-Stok
 es equations\, in particular\, a two-dimensional Taylor-Green vortex and t
 hree-dimensional Beltrami flows\, for example an Arnold-Beltrami-Childress
  flow. The simulation algorithms follow from a completely probabilistic ap
 proach.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2332@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20240314T104500
DTEND;TZID=Europe/Paris:20240314T114500
DTSTAMP:20240305T094246Z
URL:https://iecl.univ-lorraine.fr/events/toshihiro-abe/
SUMMARY:A construction of cylindrical distribution based on the normal dist
 ribution and its regression modeling
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Toshihi
 ro Abe (Hosei University) : Cylindrical distributions are joint distributi
 ons of a circular variable and a linear variable\, where the circular vari
 able affects the linear variable. In this paper\, we consider a class of c
 ylindrical distributions based on the normal distribution\, which have nor
 mal and angular conditional and marginal distributions. The distribution b
 ased on the normal distribution has the advantages of easy random number g
 eneration and simple Fisher information matrix. We also consider a regress
 ion model using the cylindrical distribution. Examples of estimation will 
 be given for real data\, and a new methodology of data analysis using the 
 cylinder model will be given. Furthermore\, we also discuss some potential
  extensions of the cylindrical distribution.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2464@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20240321T104500
DTEND;TZID=Europe/Paris:20240321T114500
DTSTAMP:20240308T085724Z
URL:https://iecl.univ-lorraine.fr/events/alexis-prevost/
SUMMARY:Exposants critiques pour le champ libre gaussien sur le système de
  câbles en dimensions intermédiaires.
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Alexis 
 Prevost (Université de Genève) : La transition de phase associée à un 
 modèle de percolation peut être quantifiée à l'aide d'un certain nombr
 e de constantes\, appelées exposants critiques\, qui décrivent la vitess
 e à laquelle certaines quantités décroissent au voisinage du point crit
 ique. J'expliquerai comment calculer certains de ces exposants critiques q
 uand le modèle de percolation est le champ libre gaussien sur le système
  de câbles en dimension trois ou quatre.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2363@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20240328T104500
DTEND;TZID=Europe/Paris:20240328T114500
DTSTAMP:20240209T140923Z
URL:https://iecl.univ-lorraine.fr/events/nicolas-chenavier-2/
SUMMARY:Maximum de vraisemblance composite pour un champ aléatoire de Brow
 n-Resnick en infill
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Nicolas
  Chenavier (Université du Littoral Côte d'Opale) : Dans cet exposé\, o
 n s'intéresse à un certain type de champ aléatoire: le champ de Brown-R
 esnick. La loi de ce dernier est décrite par deux paramètres: l'un d'éc
 helle\, l'autre de Hurst. On suppose que le champ est observé dans une f
 enêtre fixée en un nombre fini de sites. Les sites sont donnés par la r
 éalisation d'un processus ponctuel de Poisson. Estimer les paramètres pa
 r maximum de vraisemblance est en pratique impossible car les lois fini-di
 mensionnelles ne peuvent être calculées de façon efficace. Pour y remé
 dier\, nous considérons les estimateurs par maximum de vraisemblance comp
 osite en retenant comme pairs les pairs de points qui sont voisins dans l
 a triangulation de Delaunay sous-jacent et comme triplets les triplets qui
  sont sommets d'un triangle de Delaunay. Les résultats sont des théorèm
 es limites sur ces estimateurs\, lorsque l'intensité du processus de Poi
 son tend vers l'infini. Travail joint avec Christian Y. Robert.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2382@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20240404T104500
DTEND;TZID=Europe/Paris:20240404T114500
DTSTAMP:20240328T084645Z
URL:https://iecl.univ-lorraine.fr/events/jairo-cugliari/
SUMMARY:Improved linear regression prediction by transfer learning
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Jairo C
 ugliari (Université Lyon 2) : L’apprentissage par transfert (transfert 
 learning) vise à réutiliser les connaissances d'un ensemble de données 
 source vers un ensemble de données cible similaire. Alors que plusieurs 
 études abordent le problème de quoi ou comment transférer\, la question
  très importante de quand le faire reste principalement sans réponse\, s
 urtout d'un point de vue théorique pour les problèmes de régression.\nD
 ans l’exposé je présenterai le cadre général de l’apprentissage pa
 r transfert. Puis\, je détaillerai un nouveau cadre théorique pour le pr
 oblème du transfert de paramètres pour le modèle linéaire… Il est d
 émontré que la qualité du transfert pour un nouveau vecteur d'entrée d
 épend de sa représentation dans une base propre impliquant les paramètr
 es du problème. De plus\, un test statistique est construit pour prédire
  si un modèle affiné (fine tuned) a un risque quadratique de prédiction
  inférieur au modèle cible de base pour un échantillon non observé. L'
 efficacité du test est illustrée sur des données synthétiques ainsi qu
 e des données réelles de consommation d'électricité.\n\nDavid Obst\, B
 adih Ghattas\, Sandra Claudel\, Jairo Cugliari\, Yannig Goude\, Georges Op
 penheim\,\nImproved linear regression prediction by transfer learning\, CS
 DA (2022)
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle Döblin\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle Döblin:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2364@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20240411T104500
DTEND;TZID=Europe/Paris:20240411T114500
DTSTAMP:20240322T153853Z
URL:https://iecl.univ-lorraine.fr/events/mark-podolskij/
SUMMARY:On nonparametric estimation of the interaction function in particle
  system models
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Mark Po
 dolskij : This paper delves into a challenging  problem of nonparametric 
 estimation for the interaction function within diffusion-type particle sys
 tem models. We introduce two estimation methods based upon an empirical ri
 sk minimization. Our study encompasses an analysis of the stochastic and a
 pproximation errors associated with both procedures\, along with an examin
 ation of certain minimax lower bounds. In particular\, for the first metho
 d we show that there is a natural metric under which the corresponding est
 imation error of the interaction function converges to zero with parametri
 c rate which is minimax optimal. This result is rather surprising given th
 e complexity of the underlying estimation problem and rather large class o
 f interaction functions for which the above parametric rate holds.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2476@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20240418T104500
DTEND;TZID=Europe/Paris:20240418T114500
DTSTAMP:20240410T073051Z
URL:https://iecl.univ-lorraine.fr/events/delphin-senizergues/
SUMMARY:Limite d'échelle pour la limite locale de l'arbre couvrant minimal
  du graphe complet
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Delphin
  Sénizergues : Pour un graphe connexe donné muni de poids distincts sur 
 les arêtes\, il existe un unique arbre couvrant dont la somme des poids d
 es arêtes est minimale: on l'appelle l'arbre couvrant minimal. On s'inté
 resse aux propriétés asymptotiques\, pour n grand\, de l'arbre couvrant 
 minimal défini à partir du graphe complet à n sommets muni de poids i.i
 .d. sur les arêtes.\nUn résultat de convergence locale nous décrit la s
 tructure de cet objet autour d'un point typique à l'aide d'un arbre discr
 et infini. Dans un travail avec Omer Angel\, nous montrons que cet arbre i
 nfini admet une limite d'échelle: lorsqu'on fait tendre les longueurs des
  arêtes de cet arbre vers 0\, on voit apparaître un arbre continu\, dont
  on peut donner une construction explicite.\nJe présenterai les objets me
 ntionnés et expliquerai les grandes lignes de la preuve de la convergence
  du discret vers le continu.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2343@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20240516T104500
DTEND;TZID=Europe/Paris:20240516T114500
DTSTAMP:20240502T140702Z
URL:https://iecl.univ-lorraine.fr/events/marielle-simon/
SUMMARY:Quelques limites d'échelle pour le processus d'exclusion facilité
  en 1d
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Mariell
 e Simon (Université Lyon 1) : Le but de cet exposé est de présenter que
 lques résultats récents pour le processus d'exclusion facilité en une d
 imension.\nCe modèle de gaz sur réseau stochastique est soumis à de for
 tes contraintes cinétiques qui créent une transition de phase continue v
 ers un état absorbant à une valeur critique de la densité des particule
 s. Si la dynamique microscopique est symétrique\, son comportement macros
 copique (avec conditions aux limites périodiques et dans l'échelle de te
 mps diffusive)\, est régi par une EDP non linéaire appartenant aux probl
 èmes à frontières libres (ou problèmes de Stefan). L’un des ingrédi
 ents majeurs est de montrer que le système atteint la composante "ergodiq
 ue" en un temps sous-diffusif. Dans le cas asymétrique\, la densité empi
 rique converge vers l'unique solution entropique d'un problème hyperboliq
 ue de Stefan. Tous ces résultats reposent\, dans une certaine mesure\, su
 r un argument de mapping avec un processus de type zero-range\, qui ne peu
 t pas être utilisé en dimension plus grande que 1.\nD'après des travaux
  en collaboration avec O. Blondel\, H. Da Cunha\, C. Erignoux\, M. Sasada 
 et L. Zhao.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2506@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20240523T104500
DTEND;TZID=Europe/Paris:20240523T114500
DTSTAMP:20240516T115240Z
URL:https://iecl.univ-lorraine.fr/events/theo-lenoir/
SUMMARY:Graphes à décomposition modulaire prescrite\, convergence au sens
  des graphons et nombre de sous-graphe induits
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Théo L
 enoir (Polytechnique) : L'objectif de cet exposé est de montrer comment s
 e comportent certains types de modèles de graphes en particulier des mod
 èles de graphes à motifs exclus. Pour cela nous introduirons la décompo
 sition modulaire\, un outil relativement connu en algorithmique\, mais don
 t l'étude d'un point de vue probabiliste a commencé très récemment. No
 us verrons alors comment pour une large classe de modèles définies par d
 iverses contraintes sur la décomposition modulaire\, on arrive à connaî
 tre la densité de chaque graphe comme sous-graphe induit. Ce résultat im
 plique une convergence au sens des "graphons" qui peut être vue comme une
  sorte de convergence des matrices d'adjacences. On a la convergence d'un 
 graphe de taille n vers un graphe "continu" qui est appelé cographon Brow
 nien et peut être construit à partir d'une excursion Brownienne.\n\n&nbs
 p\;
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2487@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20240530T104500
DTEND;TZID=Europe/Paris:20240530T114500
DTSTAMP:20240523T071251Z
URL:https://iecl.univ-lorraine.fr/events/diyora-salimova/
SUMMARY:Deep neural network approximations for high dimensional Kolmogorov 
 PDEs
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Diyora 
 Salimova : Most of the numerical approximation methods for PDEs in the sci
 entific literature suffer from the so-called curse of dimensionality (CoD)
  in the sense that the number of computational operations and/or the numbe
 r of parameters employed in the corresponding approximation scheme grows e
 xponentially in the PDE dimension and/or the reciprocal of the desired app
 roximation precision. In recent years\, certain deep learning-based approx
 imation methods for PDEs have been proposed and various numerical simulati
 ons for such methods suggest that they might have the capacity to indeed o
 vercome the CoD in the sense that the number of real parameters used to de
 scribe the approximating neural networks grows at most polynomially in bot
 h the PDE dimension and the reciprocal of the prescribed approximation acc
 uracy. In this talk\, I will show some theoretical results which state tha
 t this is indeed the case for suitable Kolmogorov PDEs.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2522@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20240704T104500
DTEND;TZID=Europe/Paris:20240704T114500
DTSTAMP:20240625T072736Z
URL:https://iecl.univ-lorraine.fr/events/maria-dolores-jimenez-gamero-2/
SUMMARY:Testing normality of many samples
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Maria D
 olorès Jimenez Gamero (Séville) : We study the problem of simultaneously
  testing that each of k independent samples come from a normal population.
  The means and variances of those populations may differ. The proposed pro
 cedures are based on the BHEP test and they allow k to increase\, which ca
 n be even larger than the sample sizes.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2529@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris;VALUE=DATE:20240919
DTEND;TZID=Europe/Paris;VALUE=DATE:20240921
DTSTAMP:20240917T080152Z
URL:https://iecl.univ-lorraine.fr/events/reunion-dequipe-et-workshop-l2/
SUMMARY:Réunion d'équipe et workshop L2
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Simon B
 artolacci\, Victor Dubach\, Tianxiao Guo\, Aline Kurtzmann\, Ivan Nourdin 
 et Pierre Perruchaud : jeudi 19 septembre\, 10:45 réunion de rentrée de 
 l'équipe PS\n\ndu jeudi 19 septembre\, 16:00\, au vendredi 20\, 11:30 : L
 ² Workshop in Probability and Statistics à Nancy\, plus d'infos\n
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2562@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20240926T091500
DTEND;TZID=Europe/Paris:20240926T101500
DTSTAMP:20240924T081642Z
URL:https://iecl.univ-lorraine.fr/events/the-multivariate-fractional-ornst
 ein-uhlenbeck-process/
SUMMARY:The multivariate fractional Ornstein-Uhlenbeck process
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Paolo P
 igato (Roma) : In this work\, we define a multivariate version of the frac
 tional Ornstein-Uhlenbeck process\, i.e. the solution to a stochastic diff
 erential equation with affine drift and constant volatility\, driven by a 
 fractional Brownian motion. The resulting process is a multivariate statio
 nary and ergodic process\, with smoothness/regularity degree that can be d
 ifferent in each component. Such process has a richer correlation structur
 e than that of the classical diffusive case\, in the sense that the correl
 ation between i-th and j-th components is ruled by two parameters. We prop
 ose two types of estimator for these parameters\, of which we study analyt
 ically the long time asymptotic behavior\, and the finite sample behavior 
 on numerical simulations. Finally\, motivated by rough volatility modellin
 g\, we apply this framework to realized volatility time series.\n \nThis 
 is a joint work with Ranieri Dugo and Giacomo Giorgio\, based on arxiv pre
 print 2408.03051. 
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2528@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20240926T104500
DTEND;TZID=Europe/Paris:20240926T114500
DTSTAMP:20240918T133546Z
URL:https://iecl.univ-lorraine.fr/events/paul-gassiat/
SUMMARY:Un flot de gradient sur l'espace des contrôles avec condition init
 iale irrégulière
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Paul Ga
 ssiat (Paris Dauphine) : On considère un problème de contrôle consistan
 t à trouver une trajectoire reliant un point initial x à un point cible 
 y\, le système se déplaçant uniquement dans certaines directions admiss
 ibles. On suppose que les champs de vecteurs correspondants satisfont la c
 ondition de Hörmander\, de telle sorte que par un théorème classique (C
 how-Rashevskii)\, il existe des trajectoires qui satisfont cette contraint
 e. Une manière naturelle d'essayer de résoudre ce problème est via un f
 lot de gradient sur l'espace des contrôles. Cependant\, la dynamique corr
 espondante peut avoir des point-selles\, et pour obtenir un résultat de c
 onvergence il faut donc faire des hypothèses (par exemple probabilistes) 
 sur la condition initiale. Dans ce travail\, nous considérons le cas où 
 cette initialisation est irrégulière\, que nous formulons grâce à la t
 héorie des trajectoires rugueuses de Lyons. Dans des cas simples\, on pro
 uve que le flot de gradient converge vers une solution\, si la condition i
 nitiale est une trajectoire d'un mouvement Brownien (ou d'un processus de 
 régularité plus faible). La preuve combine des idées de calcul de Malli
 avin avec des inégalités de Łojasiewicz. Une motivation possible pour n
 os travaux vient de l'entraînement de réseaux de neurones résiduels pro
 fonds\, dans un régime où le nombre de paramètres par couche est fixé\
 , et la dimension du vecteur de données est élevée. Il s'agit d'un trav
 ail en collaboration avec Florin Suciu (Paris Dauphine).
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2531@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20241003T104500
DTEND;TZID=Europe/Paris:20241003T114500
DTSTAMP:20240926T184524Z
URL:https://iecl.univ-lorraine.fr/events/magalie-benefice/
SUMMARY:Couplages de processus stochastiques en géométrie sous-riemannien
 ne
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Magalie
  Bénéfice (IECL) : On s'intéresse à l'étude de couplages des mouvemen
 ts browniens sous-elliptiques sur plusieurs variétés sous-riemaniennes: 
 les groupes de Carnot libres d'ordre 2\, incluant le groupe d'Heisenberg\,
  ainsi que les groupes de matrices $SU(2)$ et $SL(2\,\\mathbb{R})$. Après
  une rapide introduction aux structures sous-Riemannienne\, nous proposero
 ns plusieurs méthodes explicites de couplages markoviens ou non markovien
 s. En particulier ces constructions mènent à des estimées du taux de co
 uplage dont on déduit des inégalités pour le semi-groupe de la chaleur 
 et pour les fonctions harmoniques que nous expliciterons.\n\nPour finir no
 us présenterons un nouveau modèle de couplage non markovien "en un coup"
  sur tous les groupes de Carnot libres de profondeur 2. Il permet notammen
 t d'obtenir des relations similaires à la formule de Bismut-Elworthy-Li p
 our les gradients de semi-groupes via l'étude d'un changement de probabil
 ité sur l'espace des vecteurs Gaussiens.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2577@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20241010T104500
DTEND;TZID=Europe/Paris:20241010T114500
DTSTAMP:20240927T133212Z
URL:https://iecl.univ-lorraine.fr/events/on-the-nonconvexity-of-push-forwa
 rd-constraints-and-its-consequences-in-machine-learning/
SUMMARY:On the nonconvexity of push-forward constraints and its consequence
 s in machine learning
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Lucas D
 e Lara (IECL) : The push-forward operation enables one to redistribute a p
 robability measure through a deterministic map. It plays a key role in sta
 tistics and optimization: many learning problems (notably from optimal tra
 nsport\, generative modeling\, and algorithmic fairness) include constrain
 ts or penalties framed as push-forward conditions on the model. However\, 
 the literature lacks general theoretical insights on the (non)convexity of
  such constraints and its consequences on the associated learning problems
 . The presented work aims at filling this gap. In a first part\, we provid
 e a range of sufficient and necessary conditions for the (non)convexity of
  two sets of functions: the maps transporting one probability measure to a
 nother\; the maps inducing equal output distributions across distinct prob
 ability measures. This highlights that for most probability measures\, the
 se push-forward constraints are not convex. In a second time\, we show how
  this result implies critical limitations on the design of convex optimiza
 tion problems for learning generative models or group-fair predictors. Thi
 s work will hopefully help researchers and practitioners have a better und
 erstanding of the critical impact of push-forward conditions onto convexit
 y.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2507@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20241017T104500
DTEND;TZID=Europe/Paris:20241017T114500
DTSTAMP:20241001T113707Z
URL:https://iecl.univ-lorraine.fr/events/maria-dolores-jimenez-gamero/
SUMMARY:Analysis of point patterns observer with errors
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Aila S
 ärkkä (Chalmers University\, Sweden) : Many natural systems are observed
  as point patterns in time\, space\, or space and time. Examples include p
 lant and cellular systems\, animal colonies\, wildfires\, and galaxies. In
  practice\, the locations of the points are not always observed correctly.
  However\, in the point process literature\, little attention has been pai
 d to the issue of errors in the location of points. In this talk\, we disc
 uss how the observed point pattern may deviate from the actual point patte
 rn\, review methods and models that exist to handle such deviations\, and 
 give some examples of data observed with errors.\n\nBased on joint work wi
 th Peter Guttorp (Norwegian Computing Center)\, Janine Illian (University 
 of Glasgow)\, Joel Kostensalo (Natural Resources Institute Finland (Luke)\
 , Mikko Kuronen (Luke)\, Mari Myllymäki (Luke)\, and Thordis Thorarinsdot
 tir (University of Oslo).
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2534@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20241107T104500
DTEND;TZID=Europe/Paris:20241107T104500
DTSTAMP:20241029T081154Z
URL:https://iecl.univ-lorraine.fr/events/andre-ribeiro/
SUMMARY:Spatio-Temporal Statistical Modelling for Environmental and Public 
 Health Applications
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 André 
 Victor Ribeiro Amaral (Imperial College London) : The increasing availabil
 ity of temporal and geo-coded data underscores the importance of spatio-te
 mporal statistical modelling in tackling complex issues across various rea
 l-world settings. In the first part of this talk\, we will briefly showcas
 e novel spatio-temporal statistical methods developed to model various typ
 es of data defined both in space and time (e.g.\, time-series\, point patt
 erns\, lattice data\, geostatistical data\, etc.)\, with a focus on applic
 ations in environmental and public health domains. In the second part\, we
  will (I) delve into the modelling of trajectory (or path) data and (II) e
 xplore the details of a statistical method for addressing spatially varyin
 g preferential sampling when modelling geostatistical data. Specifically\,
  we will account for preferential sampling by including a spatially varyin
 g coefficient that describes the dependence strength between the process t
 hat models the sampling locations and the corresponding latent field. We a
 chieve this by approximating the preferentiality component with a set of b
 asis functions\, with the corresponding coefficients estimated using the i
 ntegrated nested Laplace approximation (INLA) method. This approach allows
  for efficient inference with a low computation burden.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2532@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20241114T104500
DTEND;TZID=Europe/Paris:20241114T104500
DTSTAMP:20241104T075320Z
URL:https://iecl.univ-lorraine.fr/events/younes-zine/
SUMMARY:Hyperbolic sine-Gordon model beyond the first threshold
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Younes 
 Zine : In the past two decades\, significant progress has been made in und
 erstanding random dispersive PDEs with polynomial nonlinearities. However\
 , non-polynomial nonlinearities remain poorly understood. This talk will p
 resent recent advancements in this direction\, focusing on the well-posedn
 ess for the two-dimensional damped wave equation with a sine nonlinearity\
 , driven by additive space-time white noise.\nI will introduce the physica
 l Fourier restriction norm method\, a novel framework that addresses the c
 omplexities of non-polynomial settings. This method leverages recent devel
 opments in the Fourier restriction theory for the cone to establish crucia
 l deterministic estimates. Furthermore\, I will discuss the proof of nonli
 near smoothing for the imaginary Gaussian multiplicative chaos\, which con
 stitutes the main probabilistic component of our approach. This involves 
 examining new Feynman diagrams\, whose analysis extends beyond the classic
 al Dyson power counting criterion. This is a joint work with Tadahiro Oh (
 Edinburgh\, UK).
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle Döblin\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle Döblin:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2541@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20241121T104500
DTEND;TZID=Europe/Paris:20241121T104500
DTSTAMP:20241108T095231Z
URL:https://iecl.univ-lorraine.fr/events/guillaume-ballif/
SUMMARY:Stochastic model coupling chemical kinetics and cell population dyn
 amics
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Guillau
 me Ballif : Chemical reactions network inside cells have been extensively 
 studied in order to better understand various biological phenomena. The ma
 jority of experimental studies are performed with cells that are part of a
  growing population. This population context is rarely taken into account 
 even if selection between cells (due for example to growth) takes places w
 ithin the studied system.\nIn this talk\, I will represent such systems as
  continuous-time Markov chains. The measure-valued Markov process of the c
 ell population will take into account the chemical reactions inside the ce
 lls as well as reactions between cells. By conditioning on non-absorption\
 , we derive an equation for the expected population distribution within a 
 growing population.\nThis extension of the Chemical Master Equation provid
 es us a new framework to study cell population dynamics. I will present th
 eoretical results on long-term behaviour of the population (stationary dis
 tribution\, growth rate of the population) and an application of this fram
 ework to experimental data.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle Döblin\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle Döblin:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2586@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20241128T104500
DTEND;TZID=Europe/Paris:20241128T104500
DTSTAMP:20241115T103020Z
URL:https://iecl.univ-lorraine.fr/events/livasoa-andriamampionona/
SUMMARY:Estimation de la fonction de renouvellement sur les champs aléatoi
 res multidimensionnels
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Livasoa
  Andriamampionona (Université d’Antananarivo\, Madagascar) : \n\n\n\nLe
  processus de renouvellement fait partie des outils statistiques les plus 
 efficaces dans la théorie des files d’attente. Son espérance\, appel
 é fonction de renouvellement a été largement étudiée dans la lit
 térature. Plusieurs chercheurs ont apporté leurs contributions sur l
 ’estimation de la dite fonction. Nous présentons une nouvelle perspect
 ive dans le domaine des processus de renouvellement. Dans cette présenta
 tion\, nous étudions la convergence presque sûre et la normalité asy
 mptotique de l’estimateur de la fonction de renouvellement basée sur d
 es champs aléatoires.\n\n\n\n
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2524@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris;VALUE=DATE:20241204
DTEND;TZID=Europe/Paris;VALUE=DATE:20241207
DTSTAMP:20241125T090221Z
URL:https://iecl.univ-lorraine.fr/events/workshop-singular-spdes-invariant
 -measures-and-discrete-models/
SUMMARY:Workshop "Singular SPDEs\, invariant measures and discrete models"
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Organis
 é par Yvain Bruned : Planning\, titres et résumés ici.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2530@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20241212T104500
DTEND;TZID=Europe/Paris:20241212T114500
DTSTAMP:20241125T094944Z
URL:https://iecl.univ-lorraine.fr/events/carl-graham/
SUMMARY:Perfect simulation of the invariant laws of Markovian load-balancin
 g queueing networks
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Carl Gr
 aham (Polytechnique) : We define a wide class of Markovian load balancing 
 queueing networks\, including classic networks studied in the lively liter
 ature on the subject. Each network has identical single-server infinite-bu
 ffer queues and implements a load balancing policy to allocate each task a
 t its arrival and possibly reallocate it at service completions. The purpo
 se of the policy is to optimize server utilization under constraints such 
 as limited information\, real-time decision taking\, and network topology.
  The queue length process is not necessarily exchangeable. The invariant l
 aw is in general not known even up to normalizing constant. We provide per
 fect simulation methods in view of Monte Carlo estimation of quantities of
  interest in equilibrium\, for instance for performance evaluation. In thi
 s infinite multi-dimensional state space\, we use an unusual preorder defi
 ning an order up to permutation of the coordinates\, define a coupling in 
 which networks in this class are dominated by the network with uniform rou
 ting\, and implement dominated coupling from the past methods.\n\n[The tal
 k will be in French\, but slides will be in English.]
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2583@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20241219T104500
DTEND;TZID=Europe/Paris:20241219T114500
DTSTAMP:20241206T080617Z
URL:https://iecl.univ-lorraine.fr/events/lucas-teyssier/
SUMMARY:Temps de mélange des classes de conjugaison sans point fixe du gro
 upe symétrique
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Lucas T
 eyssier (Vancouver) : Dans cet exposé nous discuterons des temps de méla
 nge des marches aléatoires sur les graphes de Cayley du groupe symétriqu
 e dont l’ensemble générateur est une classe de conjugaison. Nous prés
 enterons comment approximer asymptotiquement certaines formules combinatoi
 res telles que la formule des équerres\, et une caractérisation du temps
  de mélange pour les classes de conjugaison sans point fixe.\nL’exposé
  repose principalement sur l’article https://arxiv.org/abs/2411.04347 r
 éalisé en collaboration avec Paul Thévenin.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2543@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20250109T104500
DTEND;TZID=Europe/Paris:20250109T114500
DTSTAMP:20250106T080758Z
URL:https://iecl.univ-lorraine.fr/events/david-dereudre/
SUMMARY:Gibbs point processes with non-summable pairwise interaction
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 David D
 ereudre (Université de Lille) : In this talk\, we discuss the question of
  Gibbs point processes in R^d with pairwise interactions that are not inte
 grable at infinity. A standard example is the Riesz potential of the form 
 g(x)=1/|x|^s where s&lt\;d. This setting has a long history\, notably beca
 use the case s=d-2 corresponds to the classical Coulomb potential\, which 
 arises from electrostatic theory. We will first address the existence of t
 he process in the infinite volume regime when a neutralizing background is
  introduced (this model is known as Jellium in theoretical physics). Subse
 quently\, we will discuss the rigidity of such point processes\, specifica
 lly hyper-uniformity and number rigidity. We will provide a state-of-the-a
 rt review and present numerous conjectures and open problems.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2644@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20250116T104500
DTEND;TZID=Europe/Paris:20250116T114500
DTSTAMP:20250106T102327Z
URL:https://iecl.univ-lorraine.fr/events/carlo-bellingeri-2/
SUMMARY:EDPSs singulières et symétries pour l'équation gKPZ à travers l
 es multi-indices
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Carlo B
 ellingeri (IECL) : \nDans cet exposé\, je présenterai les idées princip
 ales derrière la théorie des structures de régularité pour décrire le
 s solutions des EDPS singulières et comment utiliser une version plus ré
 cente de la théorie (en utilisant des multi-indices) pour calculer la dim
 ension des deux espaces associés aux symétries de l'équation de KPZ gé
 néralisée unidimensionnelle (gKPZ) : la règle de composition et l'isom
 étrie d'Itô. Notre preuve est assez élémentaire et montre que les mult
 i-indices apportent dans ce cas une simplification par rapport aux résult
 ats obtenus via les arbres décorés.  Travail en collaboration avec Y. B
 runed (Université de Lorraine).\n(Exceptionnellement\, le séminaire aura
  lieu à Metz et sera diffusé en visio en salle de conférence à Nancy.)
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de séminaires Metz\, IECL\, Metz\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Metz\, France;X-APPL
 E-RADIUS=100;X-TITLE=Salle de séminaires Metz:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2645@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20250123T104500
DTEND;TZID=Europe/Paris:20250123T114500
DTSTAMP:20250116T075513Z
URL:https://iecl.univ-lorraine.fr/events/thibault-lemoine/
SUMMARY:Partitions aléatoires\, intégrales unitaires et nombres de Hurwit
 z.
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Thibaul
 t Lemoine (Collège de France) : Dans cet exposé\, je vais décrire comme
 nt étudier un modèle probabiliste sur le groupe unitaire U(N)\, motivé 
 par la théorie de Yang–Mills\, à l'aide de la théorie des représenta
 tions. Plus précisément\, nous verrons que la fonction de partition du m
 odèle admet un développement asymptotique dont les coefficients sont li
 és aux nombres de Hurwitz\, et la démonstration passe par un couplage de
  partitions aléatoires q-uniformes. Travail en collaboration avec Mylène
  Maïda (Université de Lille).
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2671@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20250130T104500
DTEND;TZID=Europe/Paris:20250130T114500
DTSTAMP:20250120T135452Z
URL:https://iecl.univ-lorraine.fr/events/tom-guedon/
SUMMARY:Estimation de ratio de constante de normalisation: l'algorithme SAR
 IS
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Tom Gu
 édon (INRAE) : Le calcul des rapports de constante de normalisation joue 
 un rôle important dans la modélisation statistique. Deux exemples notabl
 es sont les tests d’hypothèses dans les modèles à variables latentes 
 et la comparaison de modèles en statistique bayésienne. Dans ces deux ca
 s\, le rapport de vraisemblance et le facteur de Bayes sont définis comme
  le rapport des constantes de normalisation des distributions a posteriori
 . Nous proposons dans cet article une nouvelle méthodologie qui estime ce
  rapport en utilisant le principe de l’approximation stochastique. Notre
  estimateur est consistant et asymptotiquement gaussien. Sa variance asymp
 totique est plus faible que celle de l’estimateur populaire bridge sampl
 ing. En outre\, il est beaucoup plus robuste lorsque les supports des deux
  distributions non normalisées considérées se chevauchent peu. Grâce 
 à sa définition en ligne\, notre procédure peut être intégrée dans u
 n processus d'estimation dans les modèles à variables latentes\, ce qui 
 permet ainsi de réduire l’effort de calcul. Les performances de l’est
 imateur sont illustrées par une étude de simulation et comparées à cel
 les de deux autres estimateurs : le ratio importance sampling et le bridge
  sampling.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2697@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20250206T091500
DTEND;TZID=Europe/Paris:20250206T101500
DTSTAMP:20250204T084201Z
URL:https://iecl.univ-lorraine.fr/events/efficient-estimation-for-stable-l
 evy-sdes-with-constant-scale-coefficient/
SUMMARY:Efficient estimation for stable-Lévy SDEs with constant scale coef
 ficient.
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 NGÔ Th
 ị Bảo Trâm (Université d'Évry) : In this work\, the joint parametri
 c estimation of the drift coefficient\, the scale coefficient\, and the ju
 mp activity index in stochastic differential equations driven by a symmetr
 ic stable Lévy process is considered based on high-frequency observations
 . Firstly\, the LAMN property for the corresponding Euler-type scheme is p
 roven\, and lower bounds for the estimation risk in this setting are deduc
 ed. Therefore\, when the approximation scheme experiment is asymptotically
  equivalent to the high-frequency observation of the solution of the consi
 dered stochastic differential equation\, these bounds can be transferred. 
 Secondly\, since the maximum likelihood estimator can be time-consuming fo
 r large samples\, an alternative Le Cam’s one-step procedure is proposed
  in the general setting. It is based on an initial guess estimator\, which
  is a combination of generalized variations of the trajectory for the scal
 e and the jump activity index parameters\, and a maximum likelihood type e
 stimator for the drift parameter. This proposed one-step procedure is show
 n to be fast\, asymptotically normal\, and even asymptotically efficient w
 hen the scale coefficient is constant. In addition\, the performances in t
 erms of asymptotic variance and computation time on samples of finite size
  are illustrated with simulations. This talk is based on joint work with A
 lexandre Brouste and Laurent Denis.\n\n(Exposé en français.)
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2643@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20250206T104500
DTEND;TZID=Europe/Paris:20250206T114500
DTSTAMP:20250129T092939Z
URL:https://iecl.univ-lorraine.fr/events/benoit-nieto/
SUMMARY:Multi-Mean Reverting Processes: Statistical Approaches
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Benoît
  Nieto (École Polytechnique) : Dans cette présentation\, nous nous inté
 ressons aux processus à multiples retours à la moyenne. Nous aborderons 
 l'estimation des paramètres du processus d'Ornstein-Uhlenbeck (OU)\, qui 
 présente un retour à la moyenne. Pour cela\, nous proposons un estimateu
 r basé sur les observations du supremum\, en utilisant une méthode de ps
 eudo-vraisemblance. Nous démontrerons la consistance et la normalité asy
 mptotique de cet estimateur et illustrerons son efficacité à l’aide de
  données simulées et réelles.\nNous évoquerons également brièvement 
 le processus CKLS à seuil\, qui présente plusieurs retours à la moyenne
 \, en discutant des méthodes d'estimation des paramètres de dérive et d
 e volatilité\, ainsi que des avantages d'une modélisation multi-seuils.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2694@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20250227T104500
DTEND;TZID=Europe/Paris:20250227T114500
DTSTAMP:20250219T145758Z
URL:https://iecl.univ-lorraine.fr/events/victor-trappler/
SUMMARY:Optimisation multi-objectifs en présence d'incertitudes
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Victor 
 Trappler (Lyon) : L'utilisation de modèles numériques est très répandu
 e aujourd'hui pour étudier des phénomènes physiques\, en particulier à
  des fins d'optimisation\, pour estimer des paramètres ou pour guider une
  prise de décision. Au cours de ce processus de modélisation\, il est pa
 rfois nécessaire d'ajouter des incertitudes afin de prendre en compte des
  phénomènes externes considérés comme aléatoires. Les quantités d’
 intérêt à optimiser sont donc à leur tour des variables aléatoires\, 
 et il est nécessaire d'adapter les notions d'optimalité pour prendre en 
 compte ces incertitudes.\nD'autre part\, les simulations numériques peuve
 nt représenter un coût de calcul important\, donc le nombre total d'éva
 luations du modèle est souvent limité. Ceci explique l'utilisation de mo
 dèles de substitution qui permettent de réduire le coût de calcul\, mai
 s qui peuvent aussi être utilisés pour faire de l'apprentissage actif (A
 ctive Learning) et ainsi choisir le prochain point d'évaluation du modèl
 e\, comme en optimisation bayésienne.\nDans cet exposé\, nous présenter
 ons des notions d'optimisation multi-objectifs sous incertitudes\, ainsi q
 ue des méthodes basées sur l'optimisation bayésienne pour aborder ce ge
 nre de problèmes.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2717@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20250306T091500
DTEND;TZID=Europe/Paris:20250306T101500
DTSTAMP:20250304T080209Z
URL:https://iecl.univ-lorraine.fr/events/khaydar-nurligareev/
SUMMARY:Interprétation combinatoire des coefficients dans les développeme
 nts asymptotiques
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Khaydar
  Nurligareev (Université Paris 6) : De nombreuses structures combinatoire
 s admettent\, au sens large\, une notion d’irréductibilité : les graph
 es peuvent être connexes\, les permutations indécomposables\, les polyn
 ômes irréductibles\, etc. Nous nous intéressons à la probabilité qu
 ’un tel objet pris au hasard soit irréductible\, lorsque sa taille tend
  vers l’infini. Dans cet exposé\, nous discutons de quelques méthodes 
 qui nous permettent d’obtenir les asymptotiques pour cette probabilité 
 de manière commune. Nous montrons que les coefficients apparaissant dans 
 ces asymptotiques sont entiers et qu’ils peuvent être interprétés com
 me des suites de comptage d’autres classes combinatoires “ dérivées 
 ”. De plus\, nous obtenons certaines probabilités asymptotiques qu’un
  objet combinatoire aléatoire ait un nombre donné de composantes irrédu
 ctibles. Nous appliquons notre approche aux graphes connexes\, aux graphes
  orientés fortement connexes\, aux tournois irréductibles\, aux surfaces
  à petits carreaux\, aux permutations indécomposables\, aux couplages pa
 rfaits indécomposables\, aux cartes combinatoires\, etc. Enfin\, à l’a
 ide de la théorie des espèces\, nous traitons également le modèle G(n\
 ,p) de Erdős–Rényi.\n\nCet exposé est basé sur les travaux en commun
  avec Thierry Monteil et Sergey Dovgal.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2677@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20250306T104500
DTEND;TZID=Europe/Paris:20250306T114500
DTSTAMP:20250220T153455Z
URL:https://iecl.univ-lorraine.fr/events/loic-bethencourt/
SUMMARY:Modèles cinétiques de dispersion dans des domaines et processus $
 \\alpha$-stable réfléchis
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Loïc B
 éthencourt (Université Côte d'Azur) : Dans cet exposé\, je présentera
 i quelques travaux en cours avec Nicolas Fournier. Nous nous intéressons 
 à des modèles simples décrivant le mouvement d'une particule dans un ga
 z. Une particule est alors représentée par un processus aléatoire décr
 ivant sa position et sa vitesse et nous étudions le processus de position
  lorsque le taux de collision tend vers 0. Nous nous placerons dans le cas
  où l'équilibre (en vitesse) est à queue lourdes\, et ne possède pas d
 e moment d'ordre 2. Lorsque le processus de position n'est pas restreint 
 à un domaine\, et vit dans tout l'espace\, il est assez clair que ce dern
 ier converge en loi vers un processus $\\alpha$-stable\, lorsque le taux d
 e collision tend vers 0. Nous étudions alors le cas où la particule est 
 réfléchie dans un domaine convexe de la manière suivante : lorsqu'elle 
 touche le bord du domaine\, elle est “redémarrée” avec une vitesse d
 irigée vers l'intérieur du domaine\, et distribuée selon une mesure de 
 probabilité donnée. Nous montrons que la position de la particule conver
 ge en loi vers un processus $\\alpha$-stable réfléchi dans le domaine. A
 près avoir introduit le modèle\, j'expliquerai comment nous construisons
  le processus limite en “recollant” ses excursions\, et je donnerai qu
 elques éléments de preuve concernant la convergence en loi.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2681@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris;VALUE=DATE:20250313
DTEND;TZID=Europe/Paris;VALUE=DATE:20250314
DTSTAMP:20250310T090845Z
URL:https://iecl.univ-lorraine.fr/events/journee-laurent-schwartz/
SUMMARY:Pas de séminaire : Journée Laurent Schwartz
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Page we
 b des journées : 
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Amphithéâtre 15 - Bâtiment 1er cycle\, Faculté des Sciences et
  Technologies\, Vandoeuvre-lès-Nancy\, 54506\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=Faculté des Sciences et Te
 chnologies\, Vandoeuvre-lès-Nancy\, 54506\, France;X-APPLE-RADIUS=100;X-T
 ITLE=Amphithéâtre 15 - Bâtiment 1er cycle:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2698@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20250320T091500
DTEND;TZID=Europe/Paris:20250320T101500
DTSTAMP:20250318T074958Z
URL:https://iecl.univ-lorraine.fr/events/gaspard-bernard/
SUMMARY:Testing for sphericity using spatial signs under elliptical directi
 ons
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Gaspard
  Bernard (Luxembourg) : In this talk\, we consider the problem of testing 
 for the sphericity of a collection of random vectors. It is well known tha
 t in the classical elliptical model\, testing for rotational symmetry of t
 he underlying distribution is equivalent to testing that a scatter paramet
 er is a multiple of the identity matrix. We consider the more general mode
 l of random vectors with elliptical directions introduced by R.H. Randles 
 and present a few scenarios where testing for sphericity is still equivale
 nt to testing that the scatter parameter is a multiple of the identity. Th
 ese new scenarios include\, for instance\, non-classical settings where so
 me dependence of a rather general form studied here for the first time may
  be present between observations. We study\, under these new assumptions\,
  the behavior of the classical spatial sign test and show that under certa
 in mild assumptions\, the test is asymptotically valid and has the same lo
 cal asymptotic power as in the classical elliptical scenario. We then show
  that\, contrary to some commonly held belief\, the spatial sign test enjo
 ys some local asymptotic optimality properties when it comes to testing fo
 r sphericity when the underlying distribution is strongly heavy-tailed.\n\
 n(L'exposé sera en français\, avec des slides en anglais.)
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2672@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20250320T104500
DTEND;TZID=Europe/Paris:20250320T114500
DTSTAMP:20250311T080931Z
URL:https://iecl.univ-lorraine.fr/events/marelys-crespo-navas/
SUMMARY:Stochastic Gradient Langevin Dynamics pour l'échantillonnage des d
 istributions a posteriori (faiblement) log-concaves
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Marelys
  Crespo Navas (Toulouse) : Ce travail porte sur l'étude d'un algorithme d
 e type Langevin conçu pour l'échantillonnage d'une mesure de probabilit
 é définie sur R^d qui est largement utilisé pour l'échantillonnage de 
 distributions postérieures. Cet algorithme est une version en temps conti
 nu du Stochastic Gradient Langevin Dynamics\, qui incorpore une étape d'
 échantillonnage stochastique dans la diffusion de Langevin sur amortie tr
 aditionnelle.\nL'analyse est menée dans un cadre faiblement convexe\, par
 amétré à l'aide de l'inégalité de Kurdyka–Łojasiewicz (KL). Cette 
 approche permet de traiter des configurations avec une courbure décroissa
 nte et est beaucoup moins restrictive que le cas fortement convexe.\n\nOn 
 a établi des estimations explicites non-asymptotiques du temps de simulat
 ion nécessaire pour obtenir une epsilon-approximation (en termes d'entrop
 ie). Une attention particulière est accordée au coût de calcul\, en ten
 ant compte du nombre d'observations (n) générant la distribution postér
 ieure\, de la dimension de l'espace ambiant (d) où se trouve le paramètr
 e d'intérêt\, ainsi que du paramètre r impliqué dans l'inégalité KL\
 , qui varie de 0 (cas fortement log-concave) à 1 (cas limite de Laplace).
 \n
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2693@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20250327T104500
DTEND;TZID=Europe/Paris:20250327T114500
DTSTAMP:20250319T083121Z
URL:https://iecl.univ-lorraine.fr/events/manal-jakani/
SUMMARY:Viscosity solutions for systems of variational inequalities with no
 nlinear boundary conditions on bounded domains
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Manal J
 akani (ENSAE) : We study a system of partial differential equations (PDEs)
  with interconnected obstacles and Neumann-type boundary conditions on a s
 mooth bounded domain D. This system is the Hamilton-Jacobi-Bellman system
  of equations associated with multidimensional switching problem in finite
  horizon when the state process is constrained to live in the domain D. 
 We prove the existence of a unique continuous viscosity solution. The exis
 tence of a viscosity solution is obtained using a probabilistic approach w
 hich connects the system of PDEs to a system of backward stochastic differ
 ential equations\, where randomness is constrained to stay in the domain D
 . The second main result consists in verifying the maximum principle\, whi
 ch ensures the comparison between any viscosity sub-solution and super-sol
 ution of the PDEs system. This guarantees the uniqueness and continuity of
  the solution.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2749@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20250403T091500
DTEND;TZID=Europe/Paris:20250403T101500
DTSTAMP:20250402T075047Z
URL:https://iecl.univ-lorraine.fr/events/local-expansion-properties-of-par
 acontrolled-systems/
SUMMARY:Local expansion properties of paracontrolled systems
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Nicolas
  Moench (Rennes) : \n\nIn recent years\, the theories of regularity struct
 ures and paracontrolled calculus enabled the study of singular SPDE\, prov
 iding the appropriate functional framework to handle renormalisation of th
 ese equations. While regularity structures uses local descriptions of dist
 ributions through generalized Taylor expansions\,  paracontrolled calculu
 s takes a more global approach\, utilizing harmonic analysis and the conce
 pt of paracontrolled systems. We will present a result that connects these
  two theories by introducing a family of regularity structures that captur
 es the local behavior of paracontrolled systems.\n\n
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2695@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20250403T104500
DTEND;TZID=Europe/Paris:20250403T114500
DTSTAMP:20250402T081621Z
URL:https://iecl.univ-lorraine.fr/events/yves-ngounou-bakam/
SUMMARY:Advancing Copula Methods: Nonparametric Estimation\, Smooth Testing
 \, and Data-Driven Clustering
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Yves Ng
 ounou Bakam (ENSAI) : \n\n\n\nCopulas\, introduced in the 1950’s and red
 iscovered in recent years\, are powerful tools for modeling dependence str
 uctures between multidimensional variables. These tools are particularly v
 aluable in fields like finance\, insurance\, economics\, and biol- ogy\, w
 here understanding the relationships between variables is critical. Despit
 e their generality\, copulas can present significant challenges\, particul
 arly when estimating dependence structures in complex datasets\, especiall
 y when dealing with data from different sources\, scales\, and shapes.\n\n
 This work addresses three core challenges in copula modeling: estimation\,
  testing\, and clustering. We first propose a nonparametric copula density
  estimator based on Legendre orthogonal polynomials. A nonparametric copul
 a estimator is then deduced by integration. Both estimates are based on a 
 set of moments that define the copulas\, and we’ll call them the copula 
 coefficients. Flexible modeling is possible even when copula densities may
  not exist due to the complete characterization of these coeffi- cients. A
  data-driven method is introduced to select the optimal number of copula c
 oefficients to use\, and extensive simulations show the superior performan
 ce of our approach compared to existing methods.\n\nNext\, we propose a sm
 ooth test for comparing K ≥ 2\, copulas simultaneously\, based on differ
 ences in their copula coefficients. The procedure involves a two-step data
 -driven procedure. In the first step\, the most significantly different co
 efficients are selected for all pairs of populations and the subsequent st
 ep utilizes these coefficients to identify populations that exhibit signif
 icant differences.\n\nFinally\, we use this test to develop a clustering m
 ethod that automatically identifies populations with similar dependence st
 ructures. They approaches\, implemented in the Kcop R package\, are demons
 trated through numerical studies and real-world applica- tions. This appro
 ach can be extended to the independent clustering in high dimension where 
 work is ongoing.\n\nThis is joint work with Denys Pommeret.\n\n\n\n
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2626@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20250424T104500
DTEND;TZID=Europe/Paris:20250424T114500
DTSTAMP:20250418T094545Z
URL:https://iecl.univ-lorraine.fr/events/ed-cohen/
SUMMARY:Analysing spatial point patterns on the surface of 3D shapes
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Ed Cohe
 n (Imperial College\, London) : Statistical methodology for spatial point 
 patterns has traditionally focused on Euclidean data and planar surfaces. 
 However\, with recent advances in 3D biological imaging technologies targe
 ting protein molecules on a cell’s plasma membrane\, spatial point patte
 rns are now being observed on complex shapes and manifolds whose geometry 
 must be respected for principled inference. Consequently\, there is now a 
 demand for tools that can analyse these data for important scientific stud
 ies in cellular and micro-biology. Motivated by studying the spatial distr
 ibution of LPS proteins on the surface of E-Coli\, we develop the fundamen
 tal functional summary statistics for the analysis of point patterns to ge
 neral convex bounded shapes and demonstrate how they can be used to test f
 or complete spatial randomness. We then develop their multi-type extension
 s\, together with a test for independence of the component marginal proces
 ses. To support these methods\, we introduce a plug-in estimator for the i
 ntensity of a spatial point process on a manifold. We conclude with a disc
 ussion on how these methods can readily be extended to a class of non-conv
 ex shapes. This talk will aim to provide an accessible overview of the ref
 erences below.\n\nReferences:\n\nWard\, E.A.K. Cohen\, N. M. Adams. Testin
 g for complete spatial randomness on 3-dimensional bounded convex shapes. 
 Spatial Statistics\, Vol. 41\, 2021.\n\nWard\, H. S. Battey and E. A. K. C
 ohen. Nonparametric estimation of the intensity function of a spatial poin
 t process on a Riemannian manifold. Biometrika\, Vol. 110\, 2023.\n\nKumar
 \, P. Inns\, S. Ward\, V. Lagage\, J. Wang\, R. Kaminska\, S. Uphoff\, E. 
 A. K. Cohen\, G. Mamou and C. Kleanthous. Immobile lipopolysaccharides and
  outer membrane proteins differentially segregate in growing Escherichia c
 oli. Proceedings of the National Academy of Sciences\, 122 (10)\, 2025\n\n
 Ward\, E. A. K. Cohen and N. M. Adams. Functional summary statistics and t
 esting for independence in marked point patterns on the surface of three-d
 imensional convex shapes. Spatial Statistics\, Vol. 67\, 2025
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2615@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20250515T104500
DTEND;TZID=Europe/Paris:20250515T114500
DTSTAMP:20250507T091413Z
URL:https://iecl.univ-lorraine.fr/events/bruno-ebner/
SUMMARY:Modern Perspectives on Hyperspherical Uniformity Testing: Maximal P
 rojections and Stein Characterizations
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Bruno E
 bner (Karlsruher Institut für Technologie) : \n\n\n\nWe introduce a novel
  methodology for testing uniformity on the unit hypersphere $S^{d−1}$\, 
 based on maximal projections and recent developments in Stein’s method. 
 The first framework provides a unified perspective that encompasses classi
 cal tests such as those of Rayleigh and Bingham\, while also revealing con
 nections to multivariate skewness and kurtosis. We derive the asymptotic n
 ull distribution using limit theorems for Banach space-valued stochastic p
 rocesses and employ tools from spherical harmonics theory to simulate the 
 corresponding limiting distributions. The test’s performance is analyzed
  under both contiguous and fixed alternatives\, and consistency is establi
 shed for a broad class of alternatives. Furthermore\, we present Bahadur e
 fficiency results for specific alternatives. Theoretical properties and em
 pirical power are assessed through comprehensive Monte Carlo simulations. 
 Complementing this\, we explore a second\, more recent ap- proach leveragi
 ng Stein characterizations to propose new testing procedures that extend t
 he insights of the projection-based framework.\n\nKeywords. uniformity tes
 ts\, maximal projections\, directional data\, stochastic pro- cesses in Ba
 nach spaces\, contiguous alternatives\, Monte Carlo simulations\, Stein’
 s method\n\n\n\n
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2766@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20250522T100000
DTEND;TZID=Europe/Paris:20250522T123000
DTSTAMP:20250509T071425Z
URL:https://iecl.univ-lorraine.fr/events/lancement-federation-de-recherche
 -mage/
SUMMARY:Lancement Fédération de recherche MaGE
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	  : Pas 
 de séminaire ce jeudi\, car il y a la matinée de lancement de la fédér
 ation de recherche MaGE.\n\nAu programme\, en amphi 3 :\n— café de bien
 venue à 10h\,\n— de 10h30 à 11h30 :  une présentation de ce qu’est
  une fédération de recherche du CNRS par Alessandra Sarti (Poitiers\, In
 smi)\, une présentation de MaGE et des structures qui la composent : le L
 MR à Reims\, l’IECL à Metz et Nancy\, l’IRMA à Strasbourg et le dé
 partement mathématique de l’IRIMAS à Mulhouse\,\n— à 11h30 : confé
 rence de Karin Melnick de l’Université du Luxembourg « Transformations
  conformes des variétés lorentziennes ».Lien vers l’affiche : https:
 //owncloud.math.unistra.fr/index.php/s/YzrjO3rBPTBdds5\n\n
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:2750@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20250619T104500
DTEND;TZID=Europe/Paris:20250619T114500
DTSTAMP:20250606T091638Z
URL:https://iecl.univ-lorraine.fr/events/brieuc-frenais-2/
SUMMARY:Limite hydrodynamique de processus de branchement avec sélection
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Brieuc 
 Frénais (IECL) : Les systèmes de particules sont des outils souvent util
 isés pour modéliser des populations interagissant entre elles et/ou avec
  un environnement extérieur. Je présenterai un modèle appelé N-BMP (N-
 processus de Markov branchant) dans lequel les particules ont des trajecto
 ires indépendantes sur la droite réelle\, et se séparent en deux (on pa
 rle de branchement) à des instants aléatoires successifs indépendants e
 t distribués selon une loi exponentielle. Pour garder une population de t
 aille constante au cours du temps\, la particule la plus basse est tuée 
 à chaque instant de branchement. Je poserai la question de la limite hydr
 odynamique de ce processus\, c'est-à-dire le comportement du processus ob
 tenu quand le nombre N de particules tend vers l'infini. Le cas où les pa
 rticules ont des trajectoires browniennes a notamment été étudié depui
 s 2017\, et mis en relation avec un problème à frontière libre associé
  à l'équation de la chaleur. Je présenterai les résultats que nous avo
 ns obtenus\, qui donnent un cadre plus général dans lequel le N-BMP a un
 e limite hydrodynamique\, en faisant intervenir une frontière analogue à
  celle qui apparaît dans le cas brownien. Travail en collaboration avec J
 ean Bérard.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2731@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20250626T104500
DTEND;TZID=Europe/Paris:20250626T114500
DTSTAMP:20250616T065738Z
URL:https://iecl.univ-lorraine.fr/events/jacek-wesolowski/
SUMMARY:From long random Motzkin paths to KPZ related asymptotics
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	  Jacek 
 Wesolowski (Warsaw University of technology) : We will consider Motzkin
  paths of length L with general weights on the edges and endpoint
 s and discuss limiting  behavior of their initial and final segments as L
  becomes large. We then go for macroscopic limits of the resulting proces
 ses. In two different regimes\, we obtain Markov processes which cons
 itute the non-Brownian part of the stationary measure for the KPZ e
 quation on the half-line and of conjectural stationary measure of the hypo
 thetical KPZ fixed point on the half-line. The results rely on the behavio
 r of the Al-Salam–Chihara polynomials in the neighbourhood of the right 
 end of the support  of their orthogonality measure and on Ramanujan-type 
 limiting properties of the q-Pochhammer and q-Gamma functions as q-&gt\;1.
 \n\nThis is a joint work with Wlodek Bryc (Univ. of Cincinnati) and Alexey
  Kuznetsov (York Univ. Toronto).
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2786@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20250918T104500
DTEND;TZID=Europe/Paris:20250918T114500
DTSTAMP:20250910T115509Z
URL:https://iecl.univ-lorraine.fr/events/wei-lu/
SUMMARY:Algorithmes stochastiques récursifs : d’AdaGrad aux méthodes de
  Newton stochastiques
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Wei Lu 
 : Cet exposé porte sur les algorithmes stochastiques pour le traitement s
 équentiel des données. Après un rappel des techniques classiques d’ap
 proximation stochastique\, je présenterai des contributions récentes con
 cernant le développement et l’analyse d’algorithmes avancés tels que
  Full AdaGrad et les méthodes de Newton stochastiques. Les applications i
 ncluent la régression en ligne\, l’estimation de la médiane géométri
 que et des cadres plus généraux. Nous discuterons des garanties théoriq
 ues\, des propriétés de convergence et des expériences numériques.\n\n
 Wei Lu est maître de conférences à la FST à Nancy dans notre équipe d
 epuis le 1er septembre.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2804@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20250925T104500
DTEND;TZID=Europe/Paris:20250925T114500
DTSTAMP:20250905T081312Z
URL:https://iecl.univ-lorraine.fr/events/computer-powered-chaos-in-lattice
 -models/
SUMMARY:Computer-Powered Chaos in Lattice Models
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Léo Ga
 yral (LORIA) : The study of combinatoric properties of tilings on lattice 
 models has a long history of interactions with both computability (e.g. th
 e undecidability of the domino problem) and statistical physics (e.g. the 
 Peierls argument)\, but the joining of those two interfaces is relatively 
 recent. Notably\, the question “chaotic temperature dependence” origin
 ates from the spin-glass literature\, and has been active for the last two
  decades.\n\nIn this context\, chaoticity can be summarised as the fact th
 at no converging behaviour can occur in a given model as its temperature g
 oes to 0. First formally established for an infinite spin alphabet\, this 
 property was later refined using a finite alphabet with long-range 1D inte
 ractions\, and then finite-range interactions in higher dimensions.\n\nIn 
 this talk\, I will notably focus on how the simulation of Turing machines 
 within tilings has played a key role in this evolution\, up to and includi
 ng a realisation result on the zero-temperature limit accumulation sets of
  chaotic models.\n\n(L'exposé sera en français.)
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2783@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20251002T104500
DTEND;TZID=Europe/Paris:20251002T114500
DTSTAMP:20250923T074113Z
URL:https://iecl.univ-lorraine.fr/events/matteo-dachille/
SUMMARY:Le bijou et les deux cadrans de la mosaïque de Poisson-Voronoï id
 éale
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Matteo 
 D'achille (IECL) : Nous discuterons de la limite de faible intensité d'un
 e mosaïque de Poisson-Voronoï\, alias « ideal Poisson-Voronoi tessellat
 ion » (IPVT). Dans l'espace hyperbolique de dimension d\, une simple desc
 ription poissonienne de la cellule qui contient l'origine (cellule zéro) 
 permet d'étudier des propriétés fines des tuiles de l'IPVT. Cette descr
 iption poissonienne de l'IPVT reste simple dans d'autres cas\, tels que le
  produit cartésien de plans hyperboliques.\n\nExposé basé sur un travai
 l en collaboration avec Nicolas Curien\, Nathanaël Enriquez\, Russell Lyo
 ns et Meltem Ünel (à paraître sur Ann. Probab.)\, et sur 2412.00822.\n\
 nAvec des réalisations physiques de la cellule zéro de l'IPVT de l'espac
 e hyperbolique tridimensionnel dans le modèle de la boule de Poincaré (
 « bijou »).
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2825@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20251009T104500
DTEND;TZID=Europe/Paris:20251009T114500
DTSTAMP:20251002T071032Z
URL:https://iecl.univ-lorraine.fr/events/paul-laubie/
SUMMARY:Une approche catégorique des structures de régularités
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Paul La
 ubie (IECL) : Les structures de régularités permettent de construire des
  théories de solutions afin de résoudre des équations aux dérivées pa
 rtielles stochastiques (EDPs).\nAprès une introduction aux structures de 
 régularité\, nous verrons qu'il est possible de les interpréter dans un
  cadre catégorique via la théorie des espèces combinatoires.\nNous verr
 ons ensuite deux applications de ce point de vue\, tout d'abord un théor
 ème de Y. Bruned et V. Dotsenko\, puis un préprint récent de Y. Bruned 
 et P. Laubie.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2792@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20251016T104500
DTEND;TZID=Europe/Paris:20251016T114500
DTSTAMP:20251008T094424Z
URL:https://iecl.univ-lorraine.fr/events/elie-cerf/
SUMMARY:Loi des grands nombres pour un processus de croissance-fragmentatio
 n-coagulation
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Elie Ce
 rf (IECL) : \n\n\nNous nous intéressons à une population de particules c
 aractérisées par leurs masses et évoluant selon plusieurs mécanismes: 
 une croissance déterministe selon la disponibilité d'une ressource parta
 gée\, des événements aléatoires intrinsèques à chaque particule comm
 e la fragmentation en plusieurs particules plus petites et des événement
 s aléatoires d'interactions entre deux particules comme la coagulation de
  celles-ci en une plus grosse particule.\nLe but de cet exposé est d'abor
 d de décrire l'évolution de cette population par l'introduction d'un pro
 cessus à valeurs mesures\, puis d'en étudier la convergence\, après ren
 ormalisation\, vers une trajectoire déterministe lorsque le nombre de par
 ticules dans la population initiale tend vers l'infini. En particulier\, n
 ous discuterons la généralité des hypothèses sur les paramètres du mo
 dèle nécessaires à la convergence du processus en nous concentrant sur 
 la loi de fragmentation des particules.\n\n\n
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2835@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20251106T104500
DTEND;TZID=Europe/Paris:20251106T114500
DTSTAMP:20251030T142658Z
URL:https://iecl.univ-lorraine.fr/events/irene-votsi/
SUMMARY:Risk indicators for (hidden) semi-Markov processes
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Irène 
 Votsi : The aim of this presentation is to show different results obtained
  in the field of semi-Markov processes focusing on risk/reliability indica
 tors and related statistical inference aspects. The first part concerns di
 screte-time semi-Markov processes which are defined in a finite state spac
 e. Theoretical results are obtained in terms of evaluation and statistical
  estimation of risk indicators such as the mean time to failure. The asymp
 totic behavior of the empirical estimators is studied in the case of one s
 ingle (large) trajectory. The results are illustrated on both real and sim
 ulated wind data. The second part of the presentation concerns partially o
 bserved semi-Markov chains such as the hidden Markov renewal and the hidde
 n semi-Markov chains. Statistical estimation results are presented for rel
 iability indicators such as the failure occurrence rate. In the third part
 \, we present bootstrap estimators of risk indicators when the state space
  is discrete along with posterior concentration rates of the kernel densit
 y when the state space is continuous. \n(Exposé en français avec des dia
 positives en anglais.)
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2787@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20251113T104500
DTEND;TZID=Europe/Paris:20251113T114500
DTSTAMP:20251104T125433Z
URL:https://iecl.univ-lorraine.fr/events/randolf-altmeyer/
SUMMARY:Bayesian nonparametrics for semi-linear stochastic PDEs
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Randolf
  Altmeyer : Stochastic partial differential equations (SPDEs) are a major 
 subject of current research in probability and analysis\, with rich method
 ologies for studying existence and regularity. At the same time\, SPDEs ar
 e increasingly used as statistical models for spatially and temporally str
 uctured data\, where inference requires learning unknown parameters or fun
 ctions from observations. In this talk\, we consider Bayesian inference fo
 r the reaction function in a stochastic reaction-diffusion equation\, base
 d on a single solution trajectory observed continuously in space over a fi
 xed time interval. We place a Gaussian process prior on the reaction funct
 ion and derive posterior contraction rates in a novel asymptotic regime in
  which the spatial domain grows while the observation horizon remains fixe
 d. In this setting\, the SPDE solution becomes spatially ergodic and conve
 rges to a stationary process\, which allows us to prove concentration ineq
 ualities for spatial averages of the solution. The proofs combine tools fr
 om Malliavin calculus - most notably the Clark–Ocone formula - with shar
 p bounds on the marginal densities of the SPDE.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2820@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20251120T104500
DTEND;TZID=Europe/Paris:20251120T114500
DTSTAMP:20251113T181225Z
URL:https://iecl.univ-lorraine.fr/events/vanessa-dan/
SUMMARY:Arbres de Bienaymé-Galton-Watson biconditionnés
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Vanessa
  Dan (École Polytechnique) : Le but de cet exposé sera d'étudier le com
 portement limite d’arbres de Bienaymé-Galton-Watson conditionnés à av
 oir un grand nombre de sommets\, dont un nombre fixé de feuilles ou de n
 œuds internes. Dans le premier cas\, nous obtiendrons un résultat univer
 sel quelle que soit la loi de reproduction. En revanche\, le conditionneme
 nt par le nombre de sommets et de noeuds internes donnera lieu à une dive
 rsité de comportements asymptotiques selon les propriétés de la loi de 
 reproduction\, allant de phénomènes de condensation à des structures d
 ’arbres plus allongées.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2819@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris;VALUE=DATE:20251127
DTEND;TZID=Europe/Paris;VALUE=DATE:20251128
DTSTAMP:20251125T092939Z
URL:https://iecl.univ-lorraine.fr/events/journee-ps-iecl-mage/
SUMMARY:Journée PS IECL-MaGe
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Page we
 b : Programme disponible ici.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2818@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris;VALUE=DATE:20251203
DTEND;TZID=Europe/Paris;VALUE=DATE:20251206
DTSTAMP:20250915T074309Z
URL:https://iecl.univ-lorraine.fr/events/workshop-operads-symmetries-for-q
 ft-and-singular-spdes/
SUMMARY:Workshop "Operads\, Symmetries for QFT and Singular SPDEs.
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Organis
 é par Yvain Bruned : Plus d'informations ici.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2796@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20251211T104500
DTEND;TZID=Europe/Paris:20251211T114500
DTSTAMP:20251201T094328Z
URL:https://iecl.univ-lorraine.fr/events/armand-ley/
SUMMARY:Decomposition of optimal transport plans and entropic selection on 
 the line
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Armand 
 Ley : We study the optimal transport problem on the real line with the cos
 t given by the distance\, a setting in which solutions (called optimal tra
 nsport plans) are typically non-unique. The first part of the talk present
 s a decomposition theorem: every optimal transport plan admits a unique de
 composition into components\, each acting on a specific region where the m
 ass moves forward\, moves backward\, or remains stationary. Building on th
 is structure\, the second part investigates the behaviour of an entropical
 ly regularized version of the problem as the regularization parameter tend
 s to zero. A natural candidate for the limit is constructed from our decom
 position together with a Strassen-type theorem for a strengthened stochast
 ic order. When the source and target distributions are sufficiently singul
 ar\, the entropic minimizers converge to this plan. In general\, all limit
  points satisfy a structural property known as weak multiplicativity.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2905@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20251211T140000
DTEND;TZID=Europe/Paris:20251211T150000
DTSTAMP:20251127T190503Z
URL:https://iecl.univ-lorraine.fr/events/seminaire-simba-kernel-based-test
 ing-for-single-cell-omics/
SUMMARY:Séminaire SIMBA : Kernel-based testing for single-cell omics
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Polina 
 Arsenteva (ENS Lyon) : Single-cell data yield profound insight into the co
 mplex nature of molecular feature distributions. However\, they also pose 
 statistical analysis challenges. A key challenge is the intricate geometry
  of these distributions\, which requires non-linear analysis methods. We p
 ropose a kernel-based framework for comparing conditions in single-cell ex
 periments that allows non-linear comparisons of different cell populations
 . In this talk\, I will explain how embedding the data in an infinite-dime
 nsional reproducing kernel Hilbert space (RKHS) facilitates non-linear ope
 rations on the data via linear operations in the feature space. I will pre
 sent a linear model in the RKHS and introduce a truncated kernel Hotelling
 -Lawley statistic with an associated kernel trick. This statistic has been
  shown to have an asymptotic chi-squared distribution\, which allows to qu
 antify the significance of the test results. The functionality and flexibi
 lity of the proposed approach will be demonstrated on scRNA-Seq data obtai
 ned in the context of cerebral arteries profiling. The goal of this analys
 is is to gain insight into the appearance of intracranial aneurysms.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2908@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20251218T091500
DTEND;TZID=Europe/Paris:20251218T101500
DTSTAMP:20251205T141921Z
URL:https://iecl.univ-lorraine.fr/events/william-da-silva/
SUMMARY:Hamburgers\, cheeseburgers and critical Liouville quantum gravity
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 William
  Da Silva : In a landmark paper\, Scott Sheffield introduced a famous bij
 ection\, called the "hamburger-cheeseburger" bijection\, to encode a certa
 in model of random planar maps as a certain queue model in a kitchen selli
 ng hamburgers and cheeseburgers. Under this correspondence\, natural geome
 tric observables have a nice "burger" interpretation\, for which Sheffield
  established scaling limit results. These scaling limits exhibit a phase t
 ransition in a special regime where the maps are believed to be "critical"
 . The goal of this talk is to present the analogue of these scaling limit 
 results in the critical case\, which can be thought of as the first conver
 gence result of planar maps in the universality class of critical Liouvill
 e quantum gravity\, in the so-called peanosphere sense. The talk is based 
 on joint work with Xingjian Hu\, Ellen Powell and Mo Dick Wong.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2800@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20251218T104500
DTEND;TZID=Europe/Paris:20251218T114500
DTSTAMP:20251209T150804Z
URL:https://iecl.univ-lorraine.fr/events/michel-davydov/
SUMMARY:Local-field equations and propagation of chaos
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Michel 
 Davydov : Many phenomena of interest in various applicative fields (epidem
 iology\, neuroscience\,...) can be idealized as interacting particle syste
 ms on random graphs. Various approaches have been proposed in recent years
  to develop tractable approximations of these dynamics that take the graph
  geometry and particle correlations into account. One of them\, introduced
  by Lacker\, Ramanan and Wu\, focuses on dynamics on sparse graphs and the
 ir local limits. Analogously to mean-field models on complete and dense gr
 aphs\, it is possible to establish so-called local-field equations on rand
 om trees that provide an autonomous description of the neighborhood of the
  root. In this talk\, we will give a general overview of the local-field a
 pproach\, as well as a recent result of quantitative propagation of chaos 
 in this framework.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2898@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20260108T093000
DTEND;TZID=Europe/Paris:20260108T103000
DTSTAMP:20251219T082922Z
URL:https://iecl.univ-lorraine.fr/events/nicolas-chenavier-3/
SUMMARY:Approximation par Poisson composé pour un processus beta-mélangea
 nt
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Nicolas
  Chenavier (Université du Littoral Côte d'Opale) : \n\nDans cet exposé\
 , nous présentons un résultat d'approximation par Poisson composé pour 
 un processus ponctuel\, appelé processus des excédents. Ce dernier consi
 ste en l'ensemble des points d'un processus sous-jacent en lesquels un év
 énement de type extrême apparait. La convergence associée à l'approxim
 ation est quantifiée en termes de distance de Wasserstein et permet d'ét
 udier divers problèmes d'extrêmes en géométrie aléatoire. Deux applic
 ations sont données\, la première concernant la distance au plus proche 
 voisin pour un processus non Poissonnien et le second portant sur les peti
 ts angles dans une triangulation de Delaunay. Travail joint avec M. Otto.\
 n\n
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2829@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20260108T110000
DTEND;TZID=Europe/Paris:20260108T120000
DTSTAMP:20251216T154251Z
URL:https://iecl.univ-lorraine.fr/events/pierre-olivier-goffard/
SUMMARY:Modèles stochastiques pour l’analyse des systèmes blockchain
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Pierre-
 Olivier Goffard : Une blockchain est une base de données décentralisée 
 permettant de sécuriser l’information transactionnelle entre utilisateu
 rs. Cet exposé débutera par une introduction à la technologie blockchai
 n et à ses applications en finance et en assurance. Nous présenterons en
 suite quelques modèles probabilistes simples permettant d’évaluer un s
 ystème blockchain selon trois dimensions : la sécurité\, l’efficacit
 é et la décentralisation. La sécurité désigne la résistance de la bl
 ockchain aux attaques \; nous illustrerons ce point avec l’exemple de l
 ’attaque par double dépense. L’efficacité correspond à la quantité
  de données traitée par unité de temps. Enfin\, la décentralisation re
 flète la répartition du pouvoir de décision parmi les membres du résea
 u assurant le maintien de la base de données.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2893@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20260115T104500
DTEND;TZID=Europe/Paris:20260115T114500
DTSTAMP:20260105T095136Z
URL:https://iecl.univ-lorraine.fr/events/patrick-tardivel/
SUMMARY:Le chemin des solutions de l’estimateur SLOPE (« Sorted L One Pe
 nalized Estimation »)
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Patrick
  Tardivel (Université de Bourgogne) : L’estimateur SLOPE a la particula
 rité d’avoir des composantes nulles (parcimonie) et des composantes ég
 ales en valeur absolue (appariement). Le nombre de groupes d’appariement
  dépend du paramètre de régularisation de l’estimateur. Ce paramètre
  peut être choisi comme un compromis pour obtenir un estimateur interpré
 table (en sélectionnant un petit nombre de groupes d’appariement) et pr
 écis (avec une faible erreur de prédiction). Trouver un tel compromis n
 écessite de calculer le chemin des solutions\, c’est-à-dire la fonctio
 n reliant le paramètre de régularisation à l’estimateur SLOPE. Durant
  cette présentation j’aborderai quelques résultats théoriques sur le 
 chemin des solutions du SLOPE\, j'introduirai une méthode numérique pour
  résoudre ce chemin et j'illustrerai cette méthode sur un jeu de donnée
 s réelles.\n\nCette présentation est basée sur un article\, en collabor
 ation avec Xavier Dupuis\, disponible en ligne au lien : https://proceedin
 gs.mlr.press/v238/dupuis24a.html
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2936@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20260115T140000
DTEND;TZID=Europe/Paris:20260115T150000
DTSTAMP:20251220T033730Z
URL:https://iecl.univ-lorraine.fr/events/seminaire-simba-dynamique-de-meta
 communaute-en-environnement-spatialement-heterogene/
SUMMARY:Séminaire SIMBA : Dynamique de métacommunauté en environnement s
 patialement hétérogène
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Madelei
 ne Kubasch (IEES\, Sorbonne Université) : Salle Jacques-Louis Lions (A006
 ) du bâtiment de l'Inria (en face de l'accueil)\n\nLes pratiques agricole
 s créent des habitats spatialement hétérogènes qui influencent la surv
 ie\, la dispersion et la diversité des espèces présentes. Compte tenu d
 e la crise actuelle de la biodiversité et de la nécessité de subvenir a
 ux besoins d'une population humaine croissante\, il est essentiel de conce
 voir des stratégies de gestion agricole qui concilient les objectifs de r
 endement et de conservation.\n\nDans ce travail\, nous introduisons un mod
 èle de métacommunauté qui représente la dynamique écologique de plusi
 eurs espèces en compétition dans une zone agricole. Nous partons d’un 
 processus stochastique à valeur mesure qui décrit la dynamique de métac
 ommunauté sur un réseau aléatoire formé par un nombre fini de patchs\,
  uniformément repartis dans le paysage agricole. Celui-ci converge vers l
 ’unique solution d’un système d’équations intégro-différentielle
 s\, lorsque le nombre de patchs tend vers l’infini. Dans le cas où la m
 étacommunauté ne contient que deux espèces\, nous étudions le comporte
 ment en temps long du modèle déterministe afin de déterminer quelles es
 pèces parviennent à survivre. Enfin\, nous procédons à une étude par 
 simulations pour explorer les possibilités de réconciliation des objecti
 fs de rendement et biodiversité.\n\nTravail en collaboration avec Nicolas
  Loeuille (iEES\, Sorbonne Université) et Manon Costa (IMT).
CATEGORIES:Séminaire Probabilités et Statistique
END:VEVENT
BEGIN:VEVENT
UID:2937@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20260122T140000
DTEND;TZID=Europe/Paris:20260122T150000
DTSTAMP:20251220T034334Z
URL:https://iecl.univ-lorraine.fr/events/seminaire-simba-sophie-baland/
SUMMARY:Séminaire SIMBA : Sophie Baland
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Sophie 
 Baland (IECL) : 
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2860@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20260129T104500
DTEND;TZID=Europe/Paris:20260129T114500
DTSTAMP:20260119T154011Z
URL:https://iecl.univ-lorraine.fr/events/etienne-pardoux/
SUMMARY:Loi des Grands nombres pour un modèle d’épidémies avec infecti
 vité variable et perte progressive d’immunité
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Etienne
  Pardoux (Marseille) : On considère un modèle d’épidémie où l’inf
 ectivité des infectieux dépend du temps écoulé depuis leur infection\,
  et les individus remis perdent leur immunité de façon progressive. Ces 
 concepts étaient déjà présents dans les publications de Kermack et McK
 endrick (deux épidémiologistes écossais) de 1927\, 1932 et 1933. En out
 re les fonctions d’infectivité et de perte d’immunité sont aléatoir
 es\,  i.i.d. entre les individus. On montre la convergence des quantités
  renormalisées\, quand la taille de la population tend vers l’infini\, 
 vers l’unique solution d’un système d’équations intégrales. Ce r
 ésultat a été publié récemment\, cf. ci-dessous. On esquissera une no
 uvelle démonstration\, sous une hypothèse plus faible que dans cette pub
 lication récente.\n\nRéférence : R. Forien\, G. Pang\, E.P.\, A.B. Zots
 a-Ngoufack : Stochastic epidemic models with varying infectivity and wanin
 g immunity\, Annals of Applied Probab. 35\, pp. 2175-2216\, 2025.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de séminaires Metz\, IECL\, Metz\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Metz\, France;X-APPL
 E-RADIUS=100;X-TITLE=Salle de séminaires Metz:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2904@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20260205T104500
DTEND;TZID=Europe/Paris:20260205T114500
DTSTAMP:20260129T143423Z
URL:https://iecl.univ-lorraine.fr/events/ariane-carrance/
SUMMARY:Limites d'arbres aléatoires à catastrophes locales
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Ariane 
 Carrance (Vienna) : Dans cet exposé\, je présenterai un nouveau modèle 
 d'arbres aléatoires qui généralise les arbres de Bienaymé-Galton-Watso
 n (BGW)\, en autorisant des corrélations spatiales entre les morts des in
 dividus\, à travers des "catastrophes locales". En particulier\, contrair
 ement aux arbres de BGW\, ce modèle ne satisfait plus la propriété de b
 ranchement\, ce qui rend leur analyse beaucoup plus compliquée. On peut t
 outefois montrer que\, dans le cas où les lois de reproduction et de mort
  ont des moments d'ordre 3 finis\, une forêt de tels arbres a la même li
 mite d'échelle qu'une forêt d'arbres de BGW critiques à variance finie\
 , c'est-à-dire la forêt brownienne.\n\nCes résultats sont issus d'un tr
 avail en collaboration avec Jérôme Casse et Nicolas Curien.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2861@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20260212T104500
DTEND;TZID=Europe/Paris:20260212T114500
DTSTAMP:20260202T101756Z
URL:https://iecl.univ-lorraine.fr/events/leticia-mattos/
SUMMARY:Clique packings in random graphs
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Leticia
  Mattos (Heidelberg) : We consider the question of how many edge-disjoint 
 near-maximal cliques may be found in the dense Erdős-Rényi random graph 
 $G(n\,p)$. Recently Acan and Kahn showed that the largest such family cont
 ains only $O(n^2/(\\log{n})^3)$ cliques\, with high probability\, which di
 sproved a conjecture of Alon and Spencer. We prove the corresponding lower
  bound\, $\\Omega(n^2/(\\log{n})^3)$\, by considering a random graph proce
 ss which sequentially selects and deletes near-maximal cliques. The main t
 ool in our analysis is the so-called Differential Equation Method. This is
  a joint work with Simon Griffiths.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2938@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20260212T140000
DTEND;TZID=Europe/Paris:20260212T150000
DTSTAMP:20260210T174050Z
URL:https://iecl.univ-lorraine.fr/events/seminaire-simba-vidhi-vidhi/
SUMMARY:Séminaire SIMBA : Vidhi Vidhi
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Vidhi V
 idhi (IECL) : An Introduction to the Boschloo's Test\n\nThe Boschloo's tes
 t is an exact test for comparing two binomial proportions and is closely r
 elated to Fisher’s exact test. By avoiding the conservatism inherent in 
 conditioning on fixed margins\, it can offer improved performance in small
 -sample settings. The test is constructed by using Fisher’s p-value as a
  test statistic and evaluating it under an unconditional framework. The ta
 lk will give a brief overview of the idea behind the test and its relation
 ship to classical exact methods.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2939@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20260305T104500
DTEND;TZID=Europe/Paris:20260305T114500
DTSTAMP:20260223T103826Z
URL:https://iecl.univ-lorraine.fr/events/davide-giraudo-2/
SUMMARY:Détection de rupture dans un échantillon de type Pareto via une U
 -statistique
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Davide 
 Giraudo (Strasbourg) : On cherche à déterminer si les observations d'un 
 échantillon sont issues de réalisations de variables aléatoires de mêm
 e loi de type Pareto\, c'est-à-dire dont la queue se comporte comme une p
 uissance négative fois une fonction à variations lentes. Pour cela\, nou
 s proposons un test basé sur une U-statistique construite en divisant l'
 échantillon en blocs de même taille et à regarder la différence moyenn
 e entre les estimateurs de Hill basés sur des paires de blocs. Nous fourn
 irons les garanties théoriques ainsi que pratiques de ce test. Il s'agit 
 d'un travail réalisé en collaboration avec Armelle Guillou\, disponible 
 ici  https://hal.science/IRMASTAT/hal-05424414v1.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2894@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20260312T104500
DTEND;TZID=Europe/Paris:20260312T114500
DTSTAMP:20260306T105938Z
URL:https://iecl.univ-lorraine.fr/events/mariana-olvera-cravioto/
SUMMARY:The DeGroot and Johnsen-Friedkin models on complex networks
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Mariana
  Olvera-Cravioto (Univ. North Carolina) : The DeGroot and the Johnsen-Frie
 dkin model are popular models for opinion formation on directed networks. 
 In both of them\, individuals update their opinions at each time-step by t
 aking a weighted average of their neighbors’ opinions\, either synchrono
 usly or asynchronously. Given a graph\, the synchronous DeGroot model can 
 be seen as the power iterations of a stochastic matrix\, and provided the 
 weight matrix is irreducible and aperiodic\, it is known to converge to co
 nsensus\, i.e.\, a common value that all individuals agree on. The Johnsen
 -Friedkin model allows an additional parameter which can be used to repres
 ent random external influences\, giving rise to an interactive particle sy
 stem that converges geometrically fast to a stationary distribution. This 
 talk explains how to analyze the limiting behavior of both models when the
  underlying social network is assumed to be a locally tree-like random gra
 ph (e.g.\, a configuration model\, an inhomogeneous random digraph\, or a 
 stochastic block model). This analysis allows us to understand the time to
  stationarity\, as well as characterize the limit\, in the DeGroot model\,
  or the stationary distribution\, in the Friedkin-Johnsen model\, in terms
  of the statistical properties of the underlying random graph.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2850@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20260319T104500
DTEND;TZID=Europe/Paris:20260319T114500
DTSTAMP:20260306T123824Z
URL:https://iecl.univ-lorraine.fr/events/nicolas-curien/
SUMMARY:Comment faire grossir des arbres (aléatoires)
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Nicolas
  Curien (Paris Saclay) : On y parlera de divers algorithmes pour faire gro
 ssir des arbres binaires uniformes et de leurs limites continues qui sont 
 des diffusions sur l’espace des arbres continus.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de séminaires Metz\, IECL\, Metz\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Metz\, France;X-APPL
 E-RADIUS=100;X-TITLE=Salle de séminaires Metz:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2983@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20260326T091500
DTEND;TZID=Europe/Paris:20260326T101500
DTSTAMP:20260318T080523Z
URL:https://iecl.univ-lorraine.fr/events/faicel-chamroukhi/
SUMMARY:Statistical Learning with Mixture-of-Experts: From Sparse Regulariz
 ed Estimation in High-Dimensional and Functional Settings to Frugal Distri
 buted Aggregation
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Faïcel
  Chamroukhi (Caen et Paris Saclay) : Modern statistical learning problems 
 often involve heterogeneous\, high-dimensional\, or distributed data\, rai
 sing two complementary challenges: controlling model complexity in high di
 mensions\, and enabling frugal inference under distributional constraints\
 , while preserving both structure and statistical consistency.\n\nIn this 
 talk\, I present contributions centered on mixture-of-experts (MoE) models
 \, a flexible latent variable framework with well-established approximatio
 n capabilities and learning guarantees for conditional densities.\n\nI fir
 st consider MoE models with high-dimensional predictors\, including functi
 onal data such as curves and time series\, and their training via Lasso-ty
 pe regularization in unsupervised settings\, enabling sparse and interpret
 able representations together with non-asymptotic model selection guarante
 es.\n\nI then address learning from data distributed across multiple sites
  due to storage\, computational\, or governance constraints\, and present 
 a frugal aggregation strategy for MoE models based on optimal transport\, 
 which constructs a reduced global estimator from locally trained models in
  a single communication round while preserving both model structure and st
 atistical consistency\, making it particularly well suited to large-scale 
 settings where communication is a major bottleneck.\n\nTogether\, these co
 ntributions position MoE as a unified framework for principled\, interpret
 able\, and scalable statistical learning\, with perspectives on frugal lea
 rning of generative models at the interface of modern machine learning and
  AI\, and applications in constrained industrial environments.\n\n[exposé
  en français.]
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2949@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20260326T104500
DTEND;TZID=Europe/Paris:20260326T114500
DTSTAMP:20260316T104020Z
URL:https://iecl.univ-lorraine.fr/events/sarah-jean-meyer/
SUMMARY:A construction of the sine-Gordon model above 6pi
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Sarah-J
 ean Meyer (Oxford) : I will present a construction of the Euclidean sine
 –Gordon quantum field theory using probabilistic methods. The approach f
 ormulates the model through a stochastic control problem and a weak forwar
 d–backward stochastic differential equation (FBSDE). The starting point 
 is a detailed study of the perturbation theory associated with the sine–
 Gordon potential across the full subcritical regime.  For a portion of th
 e subcritical regime extending above 6π\, the construction can be made ri
 gorous beyond perturbation theory using a stochastic control problem. I wi
 ll explain the different regimes of the model and why we currently cannot 
 construct the model in the full subcritical regime.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2974@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20260402T104500
DTEND;TZID=Europe/Paris:20260402T114500
DTSTAMP:20260324T084626Z
URL:https://iecl.univ-lorraine.fr/events/angelo-efoevi-koudou/
SUMMARY:Caractérisation des lois de Kummer par une propriété d'indépend
 ance\; copules de Lancaster.
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Angelo 
 Efoevi Koudou (IECL) : Dans la première partie\, nous parlerons d’une n
 ouvelle caractérisation des lois de Kummer par une propriété d’indép
 endance et du lien de cette caractérisation avec le fait que les lois de 
 Kummer sont les seules mesures invariantes d’un modèle discret introdui
 t en 2020 par Croydon et Sasada. Nous évoquerons aussi les connexions\, o
 bservées par Sasada et Uozumi en 2022\, entre de telles propriétés d'in
 dépendance et les transformations de Yang-Baxter.\n\nTravail en collabora
 tion avec Jacek Wesolowski  (Warsaw University of Technology)\, paru dans
  Bernoulli en 2025.\n\n&nbsp\;\n\nLa deuxième partie sera consacrée à u
 ne classe de copules appelée « copules de Lancaster ». Ce sont les copu
 les associées à des couples de variables aléatoires réelles dont la lo
 i a une densité admettant un développement en série impliquant les poly
 nômes orthogonaux relativement aux lois marginales. Des simulations montr
 ent que dans certains cas\, les tout premiers termes de la série suffisen
 t pour obtenir une bonne approximation de la copule. Cette partie de l'exp
 osé est basée sur un travail commun avec Denys Pommeret (Aix-Marseille U
 niversité) et Yves Ngounou Bakam (De Vinci Research Center\, Paris).
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:3004@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20260409T091500
DTEND;TZID=Europe/Paris:20260409T101500
DTSTAMP:20260327T145047Z
URL:https://iecl.univ-lorraine.fr/events/pierre-tarrago/
SUMMARY:Estimation spectrale en grande dimension
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Pierre 
 Tarrago (Sorbonne Université) : L'estimation spectrale consiste en l'esti
 mation des valeurs propres ou vecteurs propres de matrices bruitées. Dans
  cet exposé\, nous verrons comment la notion d'indépendance libre\, un c
 oncept issu de la théorie des opérateurs\, permet de construire des esti
 mateurs spectraux pertinents et d'en étudier la précision dans un régim
 e non-asymptotique. Si le temps le permet\, je présenterai des résultats
  relativement récents de Au\, Dahlqvist\, Gabriel et Male qui permettent 
 une application de ce principe dans le cas de l'estimation de grands graph
 es bruités.\n\nCet exposé s'appuie partiellement sur un travail en colla
 boration avec Octavio Arizmendi et Carlos Vargas Obieta.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2973@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20260409T104500
DTEND;TZID=Europe/Paris:20260409T114500
DTSTAMP:20260327T085820Z
URL:https://iecl.univ-lorraine.fr/events/mohamed-mrad/
SUMMARY:Optimisation de portefeuille et EDPS : de la composition de flots s
 tochastiques aux méthodes numériques
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Mohamed
  Mrad (Paris Nord) : Dans ce séminaire\, je m’intéresse au problème d
 e l’identification d’une utilité dynamique consistante\, qui\, par d
 éfinition\, coïncide avec la fonction valeur d’un problème d’optimi
 sation de portefeuille. Je montre que cette utilité est nécessairement s
 olution d’une équation aux dérivées partielles stochastique (EDPS) fo
 rward\, de second ordre et non linéaire.\n\nJ’établis ensuite une cara
 ctérisation explicite de cette solution en la représentant comme la comp
 osition de deux flots stochastiques bien choisis\, ce qui permet d’en ob
 tenir une résolution théorique.\n\nÀ partir de cette représentation\, 
 je propose une approche numérique nouvelle pour approximer la solution. C
 ela conduit naturellement à l’étude d’un problème plus général : 
 la composition de schémas d’Euler. Dans ce cadre\, nous établissons un
  résultat général de convergence pour de telles compositions\, applicab
 le en particulier à l’EDPS considérée. Cette approche permet ainsi de
  construire des méthodes numériques simples et efficaces\, fondées sur 
 la composition de schémas associés à deux équations différentielles s
 tochastiques appropriées.\n\nEnfin\, je présenterai quelques application
 s de cette approche\, illustrant son intérêt au-delà du cadre initial\,
  en particulier dans le domaine d’apprentissage des préférences.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:3043@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20260430T104500
DTEND;TZID=Europe/Paris:20260430T114500
DTSTAMP:20260416T071939Z
URL:https://iecl.univ-lorraine.fr/events/from-rough-to-multifractal-volati
 lity-topics-around-the-log-s-fbm-model/
SUMMARY:From rough to multifractal volatility: topics around the Log S-fBM 
 model
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Othmane
  Zarhali (Paris Dauphine) : We introduce a unified stochastic framework fo
 r modeling multiscale financial volatility based on the Log Stationary Fra
 ctional Brownian Motion (Log S-fBM) model. This construction provides a co
 ntinuous interpolation between multifractal volatility regimes and rough v
 olatility dynamics\, thereby capturing key empirical features observed in 
 financial time series. We develop a statistically robust Generalized Metho
 d of Moments (GMM) estimation procedure within the small intermittency reg
 ime. Empirical findings indicate that market indices exhibit pronounced ro
 ughness\, whereas individual assets display dynamics closer to the multifr
 actal limit which is reproduced by the Nester Stationary fractional Factor
  model we proposed. The framework of the Log S-fBM is further extended to 
 a multivariate setting\, enabling the joint modeling of correlated assets 
 through a multidimensional Log S-fBM structure. This extension preserves m
 arginal properties while incorporating cross-asset dependencies\, providin
 g a coherent explanation for the observed discrepancy between index-level 
 and single-asset volatility behavior. In addition\, we propose an efficien
 t simulation methodology for Volterra-type processes based on Random Fouri
 er Features (RFF) approximations of the kernel with a particular focus on 
 the S-fBM kernel. This approach yields improved numerical stability and co
 mputational efficiency\, supported by theoretical error bounds and empiric
 al validation. Overall\, the proposed framework offers a consistent and tr
 actable approach to linking rough volatility\, multifractal scaling\, and 
 factor-based structures\, with both theoretical and practical implications
  for financial modeling.
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2985@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20260521T104500
DTEND;TZID=Europe/Paris:20260521T114500
DTSTAMP:20260326T140129Z
URL:https://iecl.univ-lorraine.fr/events/pierre-andre-zitt/
SUMMARY:Pierre-André Zitt
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Pierre-
 André Zitt (Paris-Est Marne La Vallée) : 
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle Döblin\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle Döblin:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2984@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20260528T104500
DTEND;TZID=Europe/Paris:20260528T114500
DTSTAMP:20260209T105039Z
URL:https://iecl.univ-lorraine.fr/events/thomas-budzinski/
SUMMARY:Thomas Budzinski
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Thomas 
 Budzinski (ENS de Lyon) : 
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2992@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20260604T104500
DTEND;TZID=Europe/Paris:20260604T114500
DTSTAMP:20260210T080855Z
URL:https://iecl.univ-lorraine.fr/events/giorgos-vasdekis/
SUMMARY:Giorgos Vasdekis
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Giorgos
  Vasdekis (Newcastle University) : 
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VEVENT
UID:2897@iecl.univ-lorraine.fr
DTSTART;TZID=Europe/Paris:20260611T104500
DTEND;TZID=Europe/Paris:20260611T114500
DTSTAMP:20251118T125554Z
URL:https://iecl.univ-lorraine.fr/events/alex-podgorny/
SUMMARY:Alex Podgorny
DESCRIPTION:	\n					Séminaire Probabilités et Statistique\n			\n	 Alex Po
 dgorny (Strasbourg) : 
CATEGORIES:Séminaire Probabilités et Statistique
LOCATION:Salle de conférences Nancy\, IECL\, Nancy\, France
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=IECL\, Nancy\, France;X-APP
 LE-RADIUS=100;X-TITLE=Salle de conférences Nancy:geo:0,0
END:VEVENT
BEGIN:VTIMEZONE
TZID:Europe/Paris
X-LIC-LOCATION:Europe/Paris
BEGIN:STANDARD
DTSTART:20131027T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20140330T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20141026T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20150329T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20151025T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20160327T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20161030T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20170326T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20171029T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20180325T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20181028T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20190331T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20191027T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20200329T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20201025T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20210328T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20211031T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20220327T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20221030T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20230326T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20231029T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20240331T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20241027T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20250330T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20251026T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20260329T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
END:DAYLIGHT
END:VTIMEZONE
END:VCALENDAR