Improving Out-of-Distribution Robustness via Selective Augmentation

2022/01/02 by Huaxiu Yao, Yu Wang, Yao, Huaxiu +11 · 16 citations
Computer Science · Health Professions · #Artificial Intelligence in Healthcare #Domain Adaptation and Few-Shot Learning #FOS: Computer and information sciences #Machine Learning (cs.LG) #Machine Learning and Data Classification

paper · pdf · doi:10.48550/arxiv.2201.00299

openalex publication_date 2022/01/02 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Machine learning algorithms typically assume that training and test examples are drawn from the same distribution. However, distribution shift is a common problem in real-world applications and can cause models to perform dramatically worse at test time. In this paper, we specifically consider the problems of subpopulation shifts (e.g., imbalanced data) and domain shifts. While prior works often seek to explicitly regularize internal representations or predictors of the model to be domain invariant, we instead aim to learn invariant predictors without restricting the model's internal representations or predictors. This leads to a simple mixup-based technique which learns invariant predictors via selective augmentation called LISA. LISA selectively interpolates samples either with the same labels but different domains or with the same domain but different labels. Empirically, we study the effectiveness of LISA on nine benchmarks ranging from subpopulation shifts to domain shifts, and we find that LISA consistently outperforms other state-of-the-art methods and leads to more invariant predictors. We further analyze a linear setting and theoretically show how LISA leads to a smaller worst-group error.

Cited by

• TRoVe: Discovering Error-Inducing Static Feature Biases in Temporal Vision-Language Models
• MIMM-X: Disentangling Spurious Correlations for Medical Image Analysis
• D-GAP: Improving Out-of-Domain Robustness via Dataset-Agnostic and Gradient-Guided Augmentation in Amplitude and Pixel Spaces
• Spurious Correlation-Aware Embedding Regularization for Worst-Group Robustness
• Mitigating Spurious Correlation via Distributionally Robust Learning with Hierarchical Ambiguity Sets
• Universal Adaptive Environment Discovery
• Quantifying Dataset Similarity to Guide Transfer Learning
• Domain Generalization Under Posterior Drift
• Robust Context-Aware Object Recognition
• GroupCoOp: Group-robust Fine-tuning via Group Prompt Learning
• Temporal Generalization: A Reality Check
• FreqDebias: Towards Generalizable Deepfake Detection via Consistency-Driven Frequency Debiasing
• ACE and Diverse Generalization via Selective Disagreement
• Adapt in the Wild: Test-Time Entropy Minimization with Sharpness and Feature Regularization
• SPA++: Generalized Graph Spectral Alignment for Versatile Domain Adaptation
• Controllable Feature Whitening for Hyperparameter-Free Bias Mitigation

Related

• A Fine-Grained Analysis on Distribution Shift
• Robustness, Evaluation and Adaptation of Machine Learning Models in the Wild
• An Empirical Study on Distribution Shift Robustness From the Perspective of Pre-Training and Data Augmentation
• Robustness May be More Brittle than We Think under Different Degrees of Distribution Shifts
• An Analysis of Model Robustness across Concurrent Distribution Shifts