Domain Generalisation with Domain Augmented Supervised Contrastive Learning (Student Abstract)

TL;DR

This paper introduces a novel domain generalisation method that combines data augmentation with domain distance minimisation, improving performance on benchmarks by addressing limitations of data augmentation alone.

Contribution

It proposes a new approach integrating data augmentation and domain distance minimisation within an existing framework for better domain generalisation.

Findings

Outperforms baseline results on DG benchmarks.

Provides theoretical guarantees on learning performance.

Effectively reduces domain shift in experiments.

Abstract

Domain generalisation (DG) methods address the problem of domain shift, when there is a mismatch between the distributions of training and target domains. Data augmentation approaches have emerged as a promising alternative for DG. However, data augmentation alone is not sufficient to achieve lower generalisation errors. This project proposes a new method that combines data augmentation and domain distance minimisation to address the problems associated with data augmentation and provide a guarantee on the learning performance, under an existing framework. Empirically, our method outperforms baseline results on DG benchmarks.