Mathematical Methods of Modern Statistics 3
Méthodes Mathématiques en Statistiques Modernes 3
27 June – 1st July, 2022
|
Scientific Committee
Comité scientifique Emmanuel Candès (Stanford University) |
Organizing Committee
Comité d’organisation Malgorzata Bogdan (Wrocław University) |
|
Nowadays, in the Big Data and Artificial Intelligence era, with new technologies, statistics is at the core of numerous decisions and developments. The position of statistics in the world is increasing, the perception of statistics is changing as well, it is becoming more and more popular. Numerous statistical procedures are based on ad hoc methods supported by intensive computer assisted simulations, while statisticians and mathematicians not always know, face and confront there al issues of modern statistics. A much better understanding between statisticians and mathematicians is essential for development of both these fields. An urgent question is also: how to interacton the highest scientific level with world experts in medecine, epidemiology and genetics.
The general interaction principle is well established: statistics needs mathematics in order to have a solid base for its methodologies while mathematics (and probability, random matrix theory and harmonic analysis, in particular), benefits from statistics which poses new important challenges and, as it happened to be in the past, they may develop in beautiful and deep mathematical theories. The collaboration between two communities is crucial. Our aim is to gather in Luminy world class statisticians, who use excellent mathematics, and mathematicians who work in their area of interest. We plan to discuss recent achievements in modern statistics requiring an extensive use of deep mathematical methods and models. We would like to put emphasis on detailed exposition of mathematics behind the tools used instatistics. It is important to as precisely as possible describe mathematical status quo of the recent achievements of modern statistics and even more important, to identify mathematical challenges of the modern statistical theories, responding to the urgent needs of model selection in medecine, epidemiology and genetics. |
Actuellement, dans l’époque des Big Data et de l’Intelligence Artificielle, avec les nouvelles technologies, la statistique moderne est au coeur des plusieurs décisions et progrès. Le XXIème siècle est celui des données volumineuses et les problèmes soulevés sont nombreux. Une meilleure communication entre statisticiens et mathématiciens est essentielle au développement de ces deux disciplines et de leurs applications. Une question urgente est aussi : comment interagir au plus haut niveau scientifique avec les experts en médecine, épidémiologie et génétique.
La réciprocité perpétuelle entre statistique et mathématique est bien établie : les statistiques ont besoin des mathématiques comme solide base méthodologique alors que les mathématiques se nourissent des évolutions technologiques pour résoudre de nouveaux défis grâce à de très belles théories et nouveaux formalismes. La collaboration entre ces deux communautés devient cruciale. Notre objectif est de réunir à Luminy des statisticiens de renommée internationale, qui utilisent des mathématiques profondes, et des mathématiciens qui travaillent sur ces problèmes. Les dicussions tourneront autour des récentes avancées en statistiques modernes utilisant de manière extensive de complexes modèles mathématiques et probabilistes. Nous aimerions aborder notamment les outils mathématiques nécessaires aux statistiques modernes et d’écrire de manière très précise les principaux enjeux mathématiques issus des statistiques modernes. Cette conférence a pour objectif d’établir des relations et une meilleure compréhension de ces interactions. |
Felix Abramovich (Tel Aviv) Multiclass high-dimensional classification by sparse linear classifiers
Jean-Marc Bardet (Université Paris I) Efficient and consistent data-driven model selection for time series
Yoav Benjamini (Tel Aviv University) Replicability issues in medical research : Science, Math and Politics
Yuval Benjamini (Hebrew University of Jerusalem) Random-label models for evaluating highly multi-class classification tasks
Quentin Bertrand (Mila, Montreal) Implicit Differentiation in Non-Smooth Convex Learning
Malgorzata Bogdan (Wrocław University) Sparse Graphical Modelling via the Sorted L1-Norm
Marina Bogomolov (Technion) Testing hypotheses on a tree with FDR control for the highest resolution discoveries
Claire Boyer (Sorbonne Université) Is interpolation benign in random forest regression ?
Alain Durmus (ENS Paris-Saclay) The Kick-Kac teleportation algorithm : boost your favorite Markov Chain Monte Carlo using Kac formula
Mattéo Farné (University of Bologna) Large covariance estimation by penalized log-det minimization
Ruth Heller (Tel Aviv) Multiple testing of partial conjunction null hypotheses
Lucas Janson (Harvard) Controlled Discovery and Localization of Signals via Bayesian Linear Programming (BLiP)
Jaroslaw Harezlak (Indiana University) A Sparsity Inducing Nuclear-Norm Estimator (SpINNEr) for Matrix-Variate Regression in Brain Connectivity Analysis
Bartosz Kołodziejek (Warsaw University of Technology) Model selection in the space of Gaussian models invariant by symmetry
Yoshihiko Konno (Osaka City University) Stein’s unbiased risk estimate and an adaptive singular value shrinkage for estimation problem of low-rank matrix mean with unknown covariance matrix
Satoshi Kuriki (ISM Tokyo) Asymptotic expansion of the expected Minkowski functional for isotropic central limit random fields
Guillaume Lecue (ENSAE Palaiseau) A geometrical viewpoint on the benign overtting property of the minimum ℓ2-norm interpolant estimator in linear regression
Gérard Letac (Toulouse) Duality of exponential families and large deviations
Karim Lounici (Polytechnique Paris) Meta-Learning Representations with Contextual Linear Bandits
Ariane Marandon (Sorbonne Université) False clustering rate control in mixture models
Iqraa Meah (Sorbonne Université) Online Multiple Testing with Super Uniformity Reward
Blazej Miasojedow (Warsaw) Particle MCMC with Poisson resampling
Pierre Neuvial (Toulouse) Post hoc inference for genomics and neuroimaging
Piotr Pokarowski (University of Warsaw) Group Lasso Merger For Sparse Prediction With High-dimensional Categorical Data
Aaditya K. Ramdas (Carnegie Mellon University) Game-theoretic statistics and safe, anytime-valid inference
Wojciech Rejchel (Nicolaus Copernicus University) Model selection with high-dimensional categorical data
Etienne Roquain (UPMC Paris) FDR meets classication with BONuS
Saharon Rosset (Tel Aviv) Integrating Data Correlations into Modern Predictive Modeling
Sylvain Sardy (Génève) A phase transition for finding needles in nonlinear haystacks with LASSO artificial neural networks
Ulrike Schneider (TU Wien) The geometry of uniqueness and pattern detection in penalized and thresholded estimation
Tomasz Skalski (Polytechnique Wrocław) Pattern Recovery by SLOPE in orthogonal and asymptotic design
Weijie Su (Wharton Philadelphia) When Will You Become the Best Reviewer of Your Own Papers ? An Owner-Assisted Approach to Mechanism Design
Patrick Tardivel (Université de Bourgogne) Pattern recovery by SLOPE
Samuel Vaiter (CNRS Nice) Graph Neural Networks on Large Random Graphs
Matti Vihola (University of Jyväskylä) On the inference of hidden Markov models with weakly informative observations
Stefan Wager (Stanford) Noise-Induced Randomization in Regression Discontinuity Designs
Asaf Weinstein (Hebrew University of Jerusalem) Optimality in statistical inference for permutation invariant problems
Jonas Wallin (Lund) Gaussian WhittleMatérn fields on metric graphs
Jacek Wesołowski (Warsaw) Discrete parametric Bayesian graphical models
Olivier Wintenberger (Sorbonne Université) Stochastic Online Convex Optimization : Application to probabilistic time series forecasting
Daniel Yekutieli (Tel Aviv) Hierarchical Bayes modeling for high-dimensional linear regression
Margaux Zaffran (INRIA, Saclay) Distribution-free uncertainty quantication (for time series)
 |
 |
 |
 |