Deep Anti-Regularized Ensembles provide reliable out-of-distribution uncertainty quantification

no code implementations • 8 Apr 2023 • Antoine de Mathelin, Francois Deheeger, Mathilde Mougeot, Nicolas Vayatis

We derive a simple and practical approach to produce such ensembles, based on an original anti-regularization term penalizing small weights and a control process of the weight increase which maintains the in-distribution loss under an acceptable threshold.

Out-of-Distribution Detection regression +1

Fast and Accurate Importance Weighting for Correcting Sample Bias

1 code implementation • 9 Sep 2022 • Antoine de Mathelin, Francois Deheeger, Mathilde Mougeot, Nicolas Vayatis

Bias in datasets can be very detrimental for appropriate statistical estimation.

Discrepancy-Based Active Learning for Domain Adaptation

2 code implementations • ICLR 2022 • Antoine de Mathelin, Francois Deheeger, Mathilde Mougeot, Nicolas Vayatis

The goal of the paper is to design active learning strategies which lead to domain adaptation under an assumption of Lipschitz functions.

Active Learning Domain Adaptation

Adversarial Weighting for Domain Adaptation in Regression

2 code implementations • 15 Jun 2020 • Antoine de Mathelin, Guillaume Richard, Francois Deheeger, Mathilde Mougeot, Nicolas Vayatis

We present a novel instance-based approach to handle regression tasks in the context of supervised domain adaptation under an assumption of covariate shift.

Domain Adaptation regression