Out-of-distribution Generalization via Partial Feature Decorrelation

TL;DR

This paper introduces Partial Feature Decorrelation Learning (PFDL), a novel approach that improves out-of-distribution image classification by decorrelating features to enhance model stability across different data environments.

Contribution

The paper proposes a new PFDL algorithm that jointly optimizes feature decomposition and classification to address OOD generalization in image classification tasks.

Findings

PFDL effectively models feature correlations on synthetic data.

PFDL improves accuracy on real-world OOD datasets.

The method enhances model robustness against distribution shifts.

Abstract

Most deep-learning-based image classification methods assume that all samples are generated under an independent and identically distributed (IID) setting. However, out-of-distribution (OOD) generalization is more common in practice, which means an agnostic context distribution shift between training and testing environments. To address this problem, we present a novel Partial Feature Decorrelation Learning (PFDL) algorithm, which jointly optimizes a feature decomposition network and the target image classification model. The feature decomposition network decomposes feature embeddings into the independent and the correlated parts such that the correlations between features will be highlighted. Then, the correlated features help learn a stable feature representation by decorrelating the highlighted correlations while optimizing the image classification model. We verify the correlation modeling ability of the feature decomposition network on a synthetic dataset. The experiments on real-world datasets demonstrate that our method can improve the backbone model's accuracy on OOD image classification datasets.