GLC++: Source-Free Universal Domain Adaptation through Global-Local Clustering and Contrastive Affinity Learning

2024-03-21

arXiv.org

AI Summary

To address inaccurate recognition of known classes and ineffective discovery/separation of unknown classes in source-free universal domain adaptation (SF-UniDA), this paper proposes a global-local collaborative clustering framework. Methodologically, it introduces (i) a novel global single-class discriminative one-vs-all clustering and (ii) a local k-NN clustering robust to negative transfer; it also pioneers the integration of contrastive affinity learning into SF-UniDA to enhance fine-grained separability of unknown classes. Evaluated on VisDA under open-bias settings, our method achieves an H-score 18.6% higher than GATE; on Office-Home open-set benchmarks, it improves unknown-class clustering accuracy by 4.3% over GLC. This work breaks the conventional closed-set assumption, enabling joint discrimination and structured separation of both known and unknown classes under source-free conditions.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Adversarial Attacks & Robustness
• Search and Optimization: Local Search

Application Category

• Web Mining and Content Analysis: Normalization, clustering, classification, and summarization of Web text
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Deep neural networks often exhibit sub-optimal performance under covariate and category shifts. Source-Free Domain Adaptation (SFDA) presents a promising solution to this dilemma, yet most SFDA approaches are restricted to closed-set scenarios. In this paper, we explore Source-Free Universal Domain Adaptation (SF-UniDA) aiming to accurately classify"known"data belonging to common categories and segregate them from target-private"unknown"data. We propose a novel Global and Local Clustering (GLC) technique, which comprises an adaptive one-vs-all global clustering algorithm to discern between target classes, complemented by a local k-NN clustering strategy to mitigate negative transfer. Despite the effectiveness, the inherent closed-set source architecture leads to uniform treatment of"unknown"data, impeding the identification of distinct"unknown"categories. To address this, we evolve GLC to GLC++, integrating a contrastive affinity learning strategy. We examine the superiority of GLC and GLC++ across multiple benchmarks and category shift scenarios. Remarkably, in the most challenging open-partial-set scenarios, GLC and GLC++ surpass GATE by 16.7% and 18.6% in H-score on VisDA, respectively. GLC++ enhances the novel category clustering accuracy of GLC by 4.3% in open-set scenarios on Office-Home. Furthermore, the introduced contrastive learning strategy not only enhances GLC but also significantly facilitates existing methodologies.

Problem

• Addresses sub-optimal performance under covariate and category shifts
• Classifies known data and segregates unknown data in SF-UniDA
• Improves clustering accuracy with contrastive affinity learning

Innovation

• Global-Local Clustering for domain adaptation
• Contrastive affinity learning for unknown data
• One-vs-all global clustering algorithm