Statistical learning with Lipschitz and convex loss functions

Autor: Chinot, Geoffrey, Guillaume, Lecué, Matthieu, Lerasle

We obtain risk bounds for Empirical Risk Minimizers (ERM) and minmax Median-Of-Means (MOM) estimators based on loss functions that are both Lipschitz and convex. Results for the ERM are derived without assumptions on the outputs and under subgaussian

Externí odkaz: http://arxiv.org/abs/1810.01090

Robust sub-Gaussian estimation of a mean vector in nearly linear time

Autor: Jules Depersin, Guillaume Lecué

Publikováno v: The Annals of Statistics. 50

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=doi_________::6aed0d42de76e0c3e982faf4dd39fddb
https://doi.org/10.1214/21-aos2118

On the robustness to adversarial corruption and to heavy-tailed data of the Stahel-Donoho median of means

Autor: Jules Depersin, Guillaume Lecué

We consider median of means (MOM) versions of the Stahel–Donoho outlyingness (SDO) [ 23, 66] and of the Median Absolute Deviation (MAD) [ 30] functions to construct subgaussian estimators of a mean vector under adversarial contamination and heavy-t

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=doi_dedup___::654268c493a4daa4ce1337cde0cc5b0d
http://arxiv.org/abs/2101.09117

Robust Statistical learning with Lipschitz and convex loss functions

Autor: Guillaume Lecué, Matthieu Lerasle, Geoffrey Chinot

Publikováno v: Probability Theory and Related Fields
Probability Theory and Related Fields, Springer Verlag, 2019, ⟨10.1007/s00440-019-00931-3⟩
Probability Theory and Related Fields, 2019, ⟨10.1007/s00440-019-00931-3⟩

We obtain estimation and excess risk bounds for Empirical Risk Minimizers (ERM) and minmax Median-Of-Means (MOM) estimators based on loss functions that are both Lipschitz and convex. Results for the ERM are derived under weak assumptions on the outp

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=doi_dedup___::969e1edfbd3e9355708ee4dff2475f5c
https://hal.archives-ouvertes.fr/hal-01923033

A MOM-based ensemble method for robustness, subsampling and hyperparameter tuning

Autor: Matthieu Lerasle, Guillaume Lecué, Joon Kwon

Publikováno v: Electronic Journal of Statistics
Electronic Journal of Statistics, 2021, 15 (1), pp.1202-1227. ⟨10.1214/21-EJS1814⟩

Hyperparameters tuning and model selection are important steps in machine learning. Unfortunately, classical hyperparameter calibration and model selection procedures are sensitive to outliers and heavy-tailed data. In this work, we construct a selec

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=doi_dedup___::5ff351a32b4f66f168446c84f6c17fe4
http://arxiv.org/abs/1812.02435