Failing Loudly: An Empirical Study of Methods for Detecting Dataset Shift

TL;DR

This study evaluates methods for detecting dataset shift in machine learning, emphasizing the effectiveness of two-sample testing with pre-trained classifiers and domain-discriminating approaches for identifying and characterizing harmful shifts.

Contribution

It provides an empirical comparison of shift detection methods, highlighting the superior performance of two-sample testing combined with dimensionality reduction techniques.

Findings

Two-sample testing with pre-trained classifiers performs best across shifts.

Domain-discriminating methods help qualitatively characterize and assess shift harm.

Various dataset perturbations reveal the strengths of specific detection approaches.

Abstract

We might hope that when faced with unexpected inputs, well-designed software systems would fire off warnings. Machine learning (ML) systems, however, which depend strongly on properties of their inputs (e.g. the i.i.d. assumption), tend to fail silently. This paper explores the problem of building ML systems that fail loudly, investigating methods for detecting dataset shift, identifying exemplars that most typify the shift, and quantifying shift malignancy. We focus on several datasets and various perturbations to both covariates and label distributions with varying magnitudes and fractions of data affected. Interestingly, we show that across the dataset shifts that we explore, a two-sample-testing-based approach, using pre-trained classifiers for dimensionality reduction, performs best. Moreover, we demonstrate that domain-discriminating approaches tend to be helpful for characterizing shifts qualitatively and determining if they are harmful.