Contrast and Clustering: Learning Neighborhood Pair Representation for Source-free Domain Adaptation

TL;DR

This paper introduces a source-free domain adaptation method that leverages contrastive learning and neighborhood clustering to learn domain-invariant features without access to source data, showing superior results on standard benchmarks.

Contribution

It proposes a novel contrastive learning approach using neighborhood clustering and hard negative pairs for source-free domain adaptation, improving over existing methods.

Findings

Outperforms state-of-the-art methods on VisDA, Office-Home, and Office-31 benchmarks.

Effectively learns domain-invariant features without source data.

Demonstrates the effectiveness of neighborhood-based contrastive learning in domain adaptation.

Abstract

Unsupervised domain adaptation uses source data from different distributions to solve the problem of classifying data from unlabeled target domains. However, conventional methods require access to source data, which often raise concerns about data privacy. In this paper, we consider a more practical but challenging setting where the source domain data is unavailable and the target domain data is unlabeled. Specifically, we address the domain discrepancy problem from the perspective of contrastive learning. The key idea of our work is to learn a domain-invariant feature by 1) performing clustering directly in the original feature space with nearest neighbors; 2) constructing truly hard negative pairs by extended neighbors without introducing additional computational complexity; and 3) combining noise-contrastive estimation theory to gain computational advantage. We conduct careful ablation studies and extensive experiments on three common benchmarks: VisDA, Office-Home, and Office-31. The results demonstrate the superiority of our methods compared with other state-of-the-art works.