Clarinet: A One-step Approach Towards Budget-friendly Unsupervised Domain Adaptation

TL;DR

This paper introduces CLARINET, a novel method for budget-friendly unsupervised domain adaptation that uses complementary labels in the source domain, reducing data collection costs while maintaining high adaptation performance.

Contribution

The paper proposes CLARINET, a new adversarial network architecture for BFUDA that effectively utilizes complementary labels, a less costly alternative to true labels, for domain adaptation.

Findings

CLARINET outperforms baseline methods in experiments.

Complementary labels reduce data collection costs.

The approach maintains high adaptation accuracy with limited true-label data.

Abstract

In unsupervised domain adaptation (UDA), classifiers for the target domain are trained with massive true-label data from the source domain and unlabeled data from the target domain. However, it may be difficult to collect fully-true-label data in a source domain given a limited budget. To mitigate this problem, we consider a novel problem setting where the classifier for the target domain has to be trained with complementary-label data from the source domain and unlabeled data from the target domain named budget-friendly UDA (BFUDA). The key benefit is that it is much less costly to collect complementary-label source data (required by BFUDA) than collecting the true-label source data (required by ordinary UDA). To this end, the complementary label adversarial network (CLARINET) is proposed to solve the BFUDA problem. CLARINET maintains two deep networks simultaneously, where one focuses on classifying complementary-label source data and the other takes care of the source-to-target distributional adaptation. Experiments show that CLARINET significantly outperforms a series of competent baselines.