Unsupervised Domain Adaptation using Feature-Whitening and Consensus Loss
This addresses the problem of domain adaptation for machine learning practitioners, offering an incremental improvement over existing methods.
The paper tackles domain shift in classifiers by introducing a deep learning framework with domain alignment layers for feature whitening and a Min-Entropy Consensus loss, reporting improved state-of-the-art performances on digit classification and object recognition tasks.
A classifier trained on a dataset seldom works on other datasets obtained under different conditions due to domain shift. This problem is commonly addressed by domain adaptation methods. In this work we introduce a novel deep learning framework which unifies different paradigms in unsupervised domain adaptation. Specifically, we propose domain alignment layers which implement feature whitening for the purpose of matching source and target feature distributions. Additionally, we leverage the unlabeled target data by proposing the Min-Entropy Consensus loss, which regularizes training while avoiding the adoption of many user-defined hyper-parameters. We report results on publicly available datasets, considering both digit classification and object recognition tasks. We show that, in most of our experiments, our approach improves upon previous methods, setting new state-of-the-art performances.