MULTIYEAR PROGRAM
CONFERENCE
Meeting in Mathematical Statistics
Rencontres de Statistique Mathématique
12 – 16 December 2022
Scientific Committee & Organizing Committee
Comité scientifique & Comité d’organisation
Cristina Butucea (CREST, ENSAE, Institut Polytechnique de Paris)
Stanislav Minsker (University of Southern California)
Christophe Pouet (École Centrale de Marseille)
Vladimir Spokoiny (Humboldt University of Berlin)
The conference is focused on the analysis of complex data and of machine learning algorithms from the point of view of mathematical statistics: minimax and Bayesian approaches, asymptotic and non-asymptotic results, adaptation for estimation or testing (learning), oracle inequalities, etc. This analysis involves advanced theories in probability, optimization and computer science.
We will put an accent on two major trends in the current research:
-matrix and more general tensor models
-optimal transport theory
The interactions aim at developments of new methods and theoretical guarantees in the theory of machine learning, including areas such as deep learning, robustness, topic models, learning under constraints concerning the privacy of the individuals and the fairness of the algorithms.
La conférence est dédiée à l’analyse des données complexes et des algorithmes d’apprentissage du point vue de la statistique mathématique: approches minimax et Bayésiennes, résultats asymptotiques ou non-asymptotiques, adaptation pour l’estimation et les tests (apprentissage), inégalités oracle, etc. Cette analyse fait appel à des théories avancées de probabilités, d’optimisation et d’informatique.
Nous mettrons l’accent sur deux tendances majeures de la recherche actuelle:
-modèles matriciels et plus généralement tensoriels
-théorie du transport optimal
Les interactions visent aux développements de nouvelles méthodes et garanties théoriques pour l’apprentissage statistique dans des domaines variés comme l’apprentissage profond, les méthodes robustes, l’apprentissage sous contraintes de confidentialité des individus et d’équité des algorithmes.
LECTURES
Karim Lounici (CMAP – Ecole polytechnique)
Jonathan Niles-Weed (Courant Institute of Mathematical Sciences and Center for Data Science, New York University) – abstract –
TALKS
Yannick Baraud (Université du Luxembourg) Robust estimation in exponential families
Pierre Bellec (Rutgers University) Data-driven adjustments for regularized M-estimation in single-index models
Denis Belomestny (University of Duisburg-Essen) Bayesian methods in RL: how to be optimistic?
Tom Berrett (University of Warwick) Optimal nonparametric testing of Missing Completely At Random, and its connections to compatibility
Natalia Bochkina (University of Edinburgh) Adaptation in Bayesian inverse problems with fractional noise
Rui M. Castro (Eindhoven University of Technology) Detecting a (late) changepoint in the preferential attachment model
Julien Chhor (Harvard University) Begnin overfitting and nonparametric regression
Arnak Dalalyan (CREST-ENSAE, Paris) Graphon estimation for bipartire graphs and extensions
Patrik Gerber (Massachusetts Institute of Technology) Likelihood-free hypothesis testing
Arthur Gretton (University College London) A kernel Stein test for comparing latent variable models
Mikolaj Kasprzak (University of Luxembourg) How good is your Laplace approximation if the Bayesian posterior? Finite-sample error bounds for a variety of useful divergences
Vladimir Koltchinskii (Georgia Institute of Technology) Functional estimation in high-dimensional and infinite-dimensional models
Alexey Kroshnin (Weierstrass Institute for Applied Analysis and Stochastics) Robust k-means clustering in Hilbert and metric spaces
Gil Kur (Massachusetts Institute of Technology) Efficient minimax optimal estimators for multivariate convex regression
Sophie Langer (University of Twente) Statistical analysis of an image classification problem
Alexander Meister (University of Rostock) Nonparametric estimation under Gaussian measurement error with conditionally heteroscedastic variances
Jaouad Mourtada (CREST-ENSAE, Paris) Coding convex bodies under Gaussian noise and the Wills functional
Richard Nickl (University of Cambridge) Inference for diffusions from low frequency measurements
Dmitrii Ostrovskii (University of Southern California) Near-optimal model discrimination with non-disclosure properties
Maxim Panov (Technology Innovation Institute) Assigning topics to documents by usccessive projections
Marianna Pensky (University of Central Florida) Clustering in Diverse Multiplex Network Model
Flore Sentenac (CREST-ENSAE, Paris) Robust estimation of discrete distributions under local differential privacy
Bernhard Stankewitz (Humboldt University of Berlin) Early stopping for L^2-boosting in sparse high-dimensional linear models
Déborah Sulem (University of Oxford) Variational Bayes methods for Hawkes processes: concentration and adaptivity
Martin Wahl (Humboldt University of Berlin) Principal component analysis in infinite dimensions
SPONSORS
