Importance Weight Estimation and Generalization in Domain Adaptation under Label Shift
This work is significant for researchers and practitioners in domain adaptation, providing theoretical tools and methods to improve generalization performance when label distributions differ between source and target domains.
This paper addresses generalization in domain adaptation under label shift, proposing methods to estimate importance weights from labeled source to unlabeled target domains. The authors provide confidence bounds for these estimators and deploy them to derive generalization bounds in the unlabeled target domain.
We study generalization under labeled shift for categorical and general normed label spaces. We propose a series of methods to estimate the importance weights from labeled source to unlabeled target domain and provide confidence bounds for these estimators. We deploy these estimators and provide generalization bounds in the unlabeled target domain.