Menu
-
Le Labo
-
Équipes
-
Production scientifique
-
Activités
-
Interdisciplinarité
-
Formations
-
Ressources
Rechercher
Les événements de février 2021
Séminaire
Analyse Numérique et Calcul Scientifique
Jeudi 4 février 11:00-12:00 - Charles Dapogny - Laboratoire Jean Kuntzmann Séminaire d’Analyse Numérique et Calcul Scientifique Lieu : En ligne
Jeudi 11 février 11:00-12:00 - Havva Yoldaş - Institut Camille Jordan (ICJ) Séminaire d’Analyse Numérique et Calcul Scientifique Résumé : Analysis of a cross-diffusion model for rival gangs interaction in a city
In this talk, I will present a two-species cross-diffusion model inspired by a system of convection-diffusion equations. The latter system is derived from an agent-based model on a two-dimensional discrete lattice and proposed to simulate gang territorial development through graffiti markings. We provide a weak stability analysis on the cross-diffusion system by using entropy balance inequalities. We show that the system does not allow segregated solutions.
This is talk is based on a joint work with Alethea B. T. Barbaro (U. Delft), Nancy Rodriguez (U. Colorado Boulder) and Nicola Zamponi (U. Mannheim).Lieu : En ligne
Probabilités et statistique
Lundi 22 février 11:00-12:00 - Quentin Klopfenstein - IMB, Univ. Bourgogne Franche-Comté Séminaire PS : Implicit differentiation for fast hyperparameter selection in non-smooth convex learning Résumé : Most modern machine learning models require one hyperparameter to be chosen by the user upstream of the learning phase. Popular approaches use a grid of values on which to evaluate the performance of the model for a given criterion, one can think of grid-search or random-search which means fitting the given model for each value of the grid.
These methods have a major drawback : they scale exponentially with the number of hyperparameters. In this presentation, we will show that the hyperparameter selection problem can be cast as a bilevel optimization problem and will consider non-smooth models (such as the Lasso, the Elastic Net, the SVM).
We propose a first-order method that uses information about the gradient with respect to the hyperparameter to automatically select the best hyperparameter for a given criterion.
We will see that this method is very efficient even when the number of hyperparameters gets large.
Analyse Fonctionnelle
Mardi 23 février 13:45-15:00 - Michael Brannan - Texas A&M University Séminaire d’Analyse Fonctionnelle Mardi 23 février 16:00-17:00 - Michael Brannan - Texas A&M Complete logarithmic Sobolev inequalities and non-commutative Ricci curvature Résumé : I will give a brief introduction to the study of log-Sobolev type inequalities (LSI’s) for quantum Markov semigroups and some of their applications. In the context of classical heat semigroups on compact Riemannian manifolds, the famous Bakry-Emery theorem provides a beautiful connection between the geometry of the underlying manifold and the LSI, showing that a positive lower bound on the Ricci curvature implies an LSI for the heat semigroup. I will discuss an information-theoretic approach to obtain modified log-Sobolev inequalities based on non-positive non-commutative Ricci curvature lower bounds previously developed by Carlen and Maas. Using these tools, we are able to find new examples of quantum Markov semigroups satisfying a completely bounded version of the modified LSI, including heat semigroups on free quantum groups. This talk is based on joint work with Li Gao (TUM) and Marius Junge (UIUC).
Colloque / Journée
Fédération
Du 4 au 5 février - Ecole d’Hiver en Statistiques Besançon-Dijon Lieu : En visioconférence
Réunion
LMB
Vendredi 26 février 12:30-14:00 - Conseil de Laboratoire Lieu : En visioconférence
LMB
Lundi 1er février 12:30-14:00 - Commission Informatique (CI) Lieu : En visioconférence
Département de Mathématiques
Vendredi 12 février 12:30-14:00 - Réunion du LmB et du Département de mathématiques Lieu : En visioconférence
Soutenance
Doctorant
Mardi 9 février 14:00-16:00 - Soutenance de thèse d’Olivier ARRIGONI Lieu : UFR ST