Factorized Adversarial Networks for Unsupervised Domain Adaptation

TL;DR

This paper introduces Factorized Adversarial Networks (FAN), a novel approach for unsupervised domain adaptation in image classification that separates domain-specific and task-specific features for improved performance.

Contribution

The paper presents a new network architecture that factorizes feature space into domain-specific and task-specific parts, enabling better domain adaptation through adversarial training.

Findings

Outperforms state-of-the-art methods on benchmark datasets

Achieves significant improvements on large real-world tagging datasets

Demonstrates practical effectiveness of the proposed approach

Abstract

In this paper, we propose Factorized Adversarial Networks (FAN) to solve unsupervised domain adaptation problems for image classification tasks. Our networks map the data distribution into a latent feature space, which is factorized into a domain-specific subspace that contains domain-specific characteristics and a task-specific subspace that retains category information, for both source and target domains, respectively. Unsupervised domain adaptation is achieved by adversarial training to minimize the discrepancy between the distributions of two task-specific subspaces from source and target domains. We demonstrate that the proposed approach outperforms state-of-the-art methods on multiple benchmark datasets used in the literature for unsupervised domain adaptation. Furthermore, we collect two real-world tagging datasets that are much larger than existing benchmark datasets, and get significant improvement upon baselines, proving the practical value of our approach.