Transfer Joint Matching for Unsupervised Domain Adaptation

TLDR

Visual domain adaptation learns accurate classifiers for new domains using labeled images from old domains, yet remains challenging despite promising results. The authors argue that both feature matching and instance reweighting are essential when domain differences are large. They propose Transfer Joint Matching, which jointly matches features and reweights instances within a dimensionality‑reduction framework to produce domain‑invariant representations. Experiments demonstrate that TJM outperforms state‑of‑the‑art methods on cross‑domain image recognition tasks.

Abstract

Visual domain adaptation, which learns an accurate classifier for a new domain using labeled images from an old domain, has shown promising value in computer vision yet still been a challenging problem. Most prior works have explored two learning strategies independently for domain adaptation: feature matching and instance reweighting. In this paper, we show that both strategies are important and inevitable when the domain difference is substantially large. We therefore put forward a novel Transfer Joint Matching (TJM) approach to model them in a unified optimization problem. Specifically, TJM aims to reduce the domain difference by jointly matching the features and reweighting the instances across domains in a principled dimensionality reduction procedure, and construct new feature representation that is invariant to both the distribution difference and the irrelevant instances. Comprehensive experimental results verify that TJM can significantly outperform competitive methods for cross-domain image recognition problems.