ConstStyle: Robust Domain Generalization with Unified Style Transformation

2025-09-07

AI Summary

Domain Generalization (DG) aims to improve model generalization to unseen test domains, yet performance degrades significantly when few source domains are available or domain shifts are large. This paper proposes a Unified Style Transformation (UST) mechanism that theoretically models cross-domain mappings and employs learnable style normalization to enforce consistency between training and inference stages, thereby achieving implicit distribution alignment between source and target domains. Crucially, UST is the first method to guarantee strict consistency of style transformation across both phases, enhancing robustness of domain-invariant features. Evaluated on multiple standard DG benchmarks, UST achieves substantial improvements using only 2–3 source domains—up to +19.82% absolute accuracy gain—outperforming existing state-of-the-art approaches by a significant margin.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Adversarial Attacks & Robustness
• Application Domains: Humanities & Computational Social Science

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

Deep neural networks often suffer performance drops when test data distribution differs from training data. Domain Generalization (DG) aims to address this by focusing on domain-invariant features or augmenting data for greater diversity. However, these methods often struggle with limited training domains or significant gaps between seen (training) and unseen (test) domains. To enhance DG robustness, we hypothesize that it is essential for the model to be trained on data from domains that closely resemble unseen test domains-an inherently difficult task due to the absence of prior knowledge about the unseen domains. Accordingly, we propose ConstStyle, a novel approach that leverages a unified domain to capture domain-invariant features and bridge the domain gap with theoretical analysis. During training, all samples are mapped onto this unified domain, optimized for seen domains. During testing, unseen domain samples are projected similarly before predictions. By aligning both training and testing data within this unified domain, ConstStyle effectively reduces the impact of domain shifts, even with large domain gaps or few seen domains. Extensive experiments demonstrate that ConstStyle consistently outperforms existing methods across diverse scenarios. Notably, when only a limited number of seen domains are available, ConstStyle can boost accuracy up to 19.82% compared to the next best approach.

Problem

• Address performance drop from training-test distribution shift
• Bridge domain gap between seen and unseen domains
• Enhance robustness with limited training domains

Innovation

• Unified domain mapping for domain-invariant features
• Training and testing alignment in unified domain
• Projection optimization for seen and unseen domains