Meta-causal Learning for Single Domain Generalization

TL;DR

This paper introduces a meta-causal learning approach for single domain generalization, focusing on analyzing and reducing domain shift by simulating auxiliary domains and inferring causal factors, leading to improved model adaptation across unseen domains.

Contribution

It proposes a novel simulate-analyze-reduce paradigm and a meta-causal learning method to infer and utilize causal factors of domain shift for better generalization.

Findings

Effective in reducing domain shift in image classification benchmarks

Outperforms existing single domain generalization methods

Demonstrates robustness to various domain shifts

Abstract

Single domain generalization aims to learn a model from a single training domain (source domain) and apply it to multiple unseen test domains (target domains). Existing methods focus on expanding the distribution of the training domain to cover the target domains, but without estimating the domain shift between the source and target domains. In this paper, we propose a new learning paradigm, namely simulate-analyze-reduce, which first simulates the domain shift by building an auxiliary domain as the target domain, then learns to analyze the causes of domain shift, and finally learns to reduce the domain shift for model adaptation. Under this paradigm, we propose a meta-causal learning method to learn meta-knowledge, that is, how to infer the causes of domain shift between the auxiliary and source domains during training. We use the meta-knowledge to analyze the shift between the target and source domains during testing. Specifically, we perform multiple transformations on source data to generate the auxiliary domain, perform counterfactual inference to learn to discover the causal factors of the shift between the auxiliary and source domains, and incorporate the inferred causality into factor-aware domain alignments. Extensive experiments on several benchmarks of image classification show the effectiveness of our method.