Robust Classification under Class-Dependent Domain Shift

TL;DR

This paper introduces a method for training robust classifiers under class-dependent domain shifts, where data distribution changes are explained by a known variable, without label distribution shifts, using an information-theoretic approach.

Contribution

It proposes a new optimization framework with neural networks to handle class-dependent domain shifts, demonstrated on a toy dataset.

Findings

Method learns classifiers that generalize well to unseen domains.

Neural network-based approach effectively addresses class-dependent shifts.

Experimental results show improved robustness over baseline methods.

Abstract

Investigation of machine learning algorithms robust to changes between the training and test distributions is an active area of research. In this paper we explore a special type of dataset shift which we call class-dependent domain shift. It is characterized by the following features: the input data causally depends on the label, the shift in the data is fully explained by a known variable, the variable which controls the shift can depend on the label, there is no shift in the label distribution. We define a simple optimization problem with an information theoretic constraint and attempt to solve it with neural networks. Experiments on a toy dataset demonstrate the proposed method is able to learn robust classifiers which generalize well to unseen domains.