Out-of-Domain Generalization from a Single Source: An Uncertainty Quantification Approach

TL;DR

This paper introduces a meta-learning and adversarial augmentation framework with uncertainty quantification to improve out-of-domain generalization from a single source domain, demonstrating superior results on benchmarks.

Contribution

It proposes a novel approach combining adversarial domain augmentation, meta-learning, and uncertainty quantification to address single-source out-of-domain generalization with theoretical guarantees.

Findings

Outperforms existing methods on multiple benchmarks.

Effectively generates challenging domain samples.

Enhances model robustness to unseen domains.

Abstract

We are concerned with a worst-case scenario in model generalization, in the sense that a model aims to perform well on many unseen domains while there is only one single domain available for training. We propose Meta-Learning based Adversarial Domain Augmentation to solve this Out-of-Domain generalization problem. The key idea is to leverage adversarial training to create "fictitious" yet "challenging" populations, from which a model can learn to generalize with theoretical guarantees. To facilitate fast and desirable domain augmentation, we cast the model training in a meta-learning scheme and use a Wasserstein Auto-Encoder to relax the widely used worst-case constraint. We further improve our method by integrating uncertainty quantification for efficient domain generalization. Extensive experiments on multiple benchmark datasets indicate its superior performance in tackling single domain generalization.