Domain Discrepancy Measure for Complex Models in Unsupervised Domain Adaptation
This work addresses a key bottleneck in domain adaptation for researchers and practitioners, offering a more practical and scalable solution, though it is incremental in improving existing measures.
The paper tackles the problem of evaluating domain discrepancy in unsupervised domain adaptation for complex models like deep neural networks, proposing a novel measure called paired hypotheses discrepancy (PHD) that is computationally efficient, applicable to multi-class classification, and theoretically effective, with experimental validation.
Appropriately evaluating the discrepancy between domains is essential for the success of unsupervised domain adaptation. In this paper, we first point out that existing discrepancy measures are less informative when complex models such as deep neural networks are used, in addition to the facts that they can be computationally highly demanding and their range of applications is limited only to binary classification. We then propose a novel domain discrepancy measure, called the paired hypotheses discrepancy (PHD), to overcome these shortcomings. PHD is computationally efficient and applicable to multi-class classification. Through generalization error bound analysis, we theoretically show that PHD is effective even for complex models. Finally, we demonstrate the practical usefulness of PHD through experiments.