Distributionally Robust Safe Sample Elimination under Covariate Shift

TL;DR

This paper introduces DRSSS, a method combining distributionally robust optimization and safe sample screening to create models that perform consistently across different environments under covariate shift, reducing costs.

Contribution

The paper proposes a novel DRSSS approach that ensures model robustness across environments while minimizing training data and computational costs.

Findings

DRSSS achieves consistent performance across environments.

The method reduces training data requirements.

Experimental results validate effectiveness under covariate shift.

Abstract

We consider a machine learning setup where one training dataset is used to train multiple models across slightly different data distributions. This occurs when customized models are needed for various deployment environments. To reduce storage and training costs, we propose the DRSSS method, which combines distributionally robust (DR) optimization and safe sample screening (SSS). The key benefit of this method is that models trained on the reduced dataset will perform the same as those trained on the full dataset for all possible different environments. In this paper, we focus on covariate shift as a type of data distribution change and demonstrate the effectiveness of our method through experiments.