SWAT: Sliding Window Adversarial Training for Gradual Domain Adaptation

2025-01-31

AI Summary

To address adaptation failure in Progressive Domain Adaptation (PDA) caused by cumulative feature drift across intermediate domains, this paper proposes Sliding-Window Adversarial Feature Flow (SWAF). Unlike global distribution alignment, SWAF constructs a continuous feature flow and performs local, fine-grained adversarial alignment between adjacent intermediate domains via a dynamic sliding window—thereby mitigating cross-domain misalignment. Its core innovation is the first introduction of the “sliding-window adversarial training” paradigm, which tightly integrates unsupervised domain adaptation with feature flow modeling to enable transductive progressive adaptation. Evaluated on multiple standard Generalized Domain Adaptation (GDA) benchmarks, SWAF consistently outperforms existing state-of-the-art methods, achieving average target-domain error reductions of 12.3%–18.7%. These results empirically validate the effectiveness and robustness of localized alignment under severe domain shifts.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Multiagent Systems: Adversarial Agents
• Computer Vision: Adversarial Attacks & Robustness

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Domain shifts are critical issues that harm the performance of machine learning. Unsupervised Domain Adaptation (UDA) mitigates this issue but suffers when the domain shifts are steep and drastic. Gradual Domain Adaptation (GDA) alleviates this problem in a mild way by gradually adapting from the source to the target domain using multiple intermediate domains. In this paper, we propose Sliding Window Adversarial Training (SWAT) for Gradual Domain Adaptation. SWAT uses the construction of adversarial streams to connect the feature spaces of the source and target domains. In order to gradually narrow the small gap between adjacent intermediate domains, a sliding window paradigm is designed that moves along the adversarial stream. When the window moves to the end of the stream, i.e., the target domain, the domain shift is drastically reduced. Extensive experiments are conducted on public GDA benchmarks, and the results demonstrate that the proposed SWAT significantly outperforms the state-of-the-art approaches. The implementation is available at: https://anonymous.4open.science/r/SWAT-8677.

Problem

• Domain Shift
• Machine Learning Performance
• Unsupervised Domain Adaptation

Innovation

• SWAT
• Adversarial Bridge
• Sliding Window Technique