Unsupervised Domain Adaptation via Calibrating Uncertainties
It addresses domain shift for machine learning models, but appears incremental as it builds on existing uncertainty-based methods.
The paper tackles unsupervised domain adaptation by calibrating predictive uncertainties, using Renyi entropy regularization and variational Bayes, and demonstrates effectiveness on three tasks.
Unsupervised domain adaptation (UDA) aims at inferring class labels for unlabeled target domain given a related labeled source dataset. Intuitively, a model trained on source domain normally produces higher uncertainties for unseen data. In this work, we build on this assumption and propose to adapt from source to target domain via calibrating their predictive uncertainties. The uncertainty is quantified as the Renyi entropy, from which we propose a general Renyi entropy regularization (RER) framework. We further employ variational Bayes learning for reliable uncertainty estimation. In addition, calibrating the sample variance of network parameters serves as a plug-in regularizer for training. We discuss the theoretical properties of the proposed method and demonstrate its effectiveness on three domain-adaptation tasks.