Decision Boundary Optimization-Informed Domain Adaptation

2025-02-10

AI Summary

Existing MMD-based domain adaptation methods solely align source and target data distributions, neglecting optimization of the classifier’s decision boundary—thus failing to reduce the theoretical upper bound on classification error. To address this, we propose Decision Boundary-aware MMD (DB-MMD), the first method to incorporate classifier gradient information into the kernel mean embedding framework, enabling joint optimization of distribution alignment and decision boundary refinement. We rigorously derive its generalization error bound. Our approach introduces differentiable decision boundary regularization and gradient-coupled modeling, and is compatible with mainstream frameworks such as MEDA and DGA-DA. Evaluated on eight standard unsupervised domain adaptation benchmarks, DB-MMD achieves up to a 9.5 percentage-point improvement over conventional MMD baselines. It significantly enhances transfer robustness and provides stronger theoretical guarantees for domain adaptation.

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
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Maximum Mean Discrepancy (MMD) is widely used in a number of domain adaptation (DA) methods and shows its effectiveness in aligning data distributions across domains. However, in previous DA research, MMD-based DA methods focus mostly on distribution alignment, and ignore to optimize the decision boundary for classification-aware DA, thereby falling short in reducing the DA upper error bound. In this paper, we propose a strengthened MMD measurement, namely, Decision Boundary optimization-informed MMD (DB-MMD), which enables MMD to carefully take into account the decision boundaries, thereby simultaneously optimizing the distribution alignment and cross-domain classifier within a hybrid framework, and leading to a theoretical bound guided DA. We further seamlessly embed the proposed DB-MMD measurement into several popular DA methods, e.g., MEDA, DGA-DA, to demonstrate its effectiveness w.r.t different experimental settings. We carry out comprehensive experiments using 8 standard DA datasets. The experimental results show that the DB-MMD enforced DA methods improve their baseline models using plain vanilla MMD, with a margin that can be as high as 9.5.

Problem

• Optimize decision boundary for classification-aware DA
• Strengthen MMD measurement for distribution alignment
• Embed DB-MMD into popular DA methods

Innovation

• Decision Boundary optimization-informed MMD
• Simultaneously optimizes distribution alignment
• Seamlessly embedded into popular DA methods