An Analysis of Model Robustness across Concurrent Distribution Shifts

TL;DR

This study investigates how multiple simultaneous distribution shifts affect machine learning model robustness, revealing that such shifts generally degrade performance but heuristic augmentations can improve generalization across complex real-world scenarios.

Contribution

It extends prior work by analyzing models under multiple concurrent distribution shifts, including complex real-world scenarios, and evaluates a wide range of algorithms on numerous dataset pairs.

Findings

Concurrent distribution shifts usually reduce model performance.

Improvement in one shift often correlates with better performance on others.

Heuristic data augmentations outperform other methods overall.

Abstract

Machine learning models, meticulously optimized for source data, often fail to predict target data when faced with distribution shifts (DSs). Previous benchmarking studies, though extensive, have mainly focused on simple DSs. Recognizing that DSs often occur in more complex forms in real-world scenarios, we broadened our study to include multiple concurrent shifts, such as unseen domain shifts combined with spurious correlations. We evaluated 26 algorithms that range from simple heuristic augmentations to zero-shot inference using foundation models, across 168 source-target pairs from eight datasets. Our analysis of over 100K models reveals that (i) concurrent DSs typically worsen performance compared to a single shift, with certain exceptions, (ii) if a model improves generalization for one distribution shift, it tends to be effective for others, and (iii) heuristic data augmentations achieve the best overall performance on both synthetic and real-world datasets.