Aggregation of Disentanglement: Reconsidering Domain Variations in Domain Generalization

Abstract

Domain Generalization (DG) is a fundamental challenge for machine learning models, which aims to improve model generalization on various domains. Previous methods focus on generating domain invariant features from various source domains. However, we argue that the domain variantions also contain useful information, i.e., classificationaware information, for downstream tasks, which has been largely ignored. Different from learning domain invariant features from source domains, we decouple the input images into Domain Expert Features and noise. The proposed domain expert features lie in a learned latent space where the images in each domain can be classified independently, enabling the implicit use of classification-aware domain variations. Based on the analysis, we proposed a novel paradigm called Domain Disentanglement Network (DDN) to disentangle the domain expert features from the source domain images and aggregate the source domain expert features for representing the target test domain. We also propound a new contrastive learning method to guide the domain expert features to form a more balanced and separable feature space. Experiments on the widely-used benchmarks of PACS, VLCS, OfficeHome, DomainNet, and TerraIncognita demonstrate the competitive performance of our method compared to the recently proposed alternatives.

Enhancing Evolving Domain Generalization through Dynamic Latent Representations

“…A rich literature is dedicated to addressing the OOD generalization challenge, which often involves introducing additional regularizations to Empirical Risk Minimization (ERM) (Vapnik 1991) Besides, some methods are proposed to utilize domaindependent features for better OOD generalization ability, e.g., (Chattopadhyay, Balaji, and Hoffman 2020;Zhang et al 2021;Bui et al 2021;Zhang et al 2023). Yet, they still treat domain index as a discrete variable, and can not learn the evolving drift across the domains.…”

Bootstrap The Original Latent: Learning a Private Model from a Black-box Model

“…Under the so-called Model-as-a-Service (MaaS) [24] scenario, users can adapt their data via zero-shot or black-box source-free domain adaptation. These techniques work well on language tasks but are inefficient in visual tasks because the distribution gap in vision tasks are large [35]. Even foundation vision models cannot generalize on such varieties of data.…”