Active-Transductive Learning with Label-Adapted Kernels

This paper presents an efficient active-transductive approach for classification. A common approach of active learning algorithms is to focus on querying points near the class boundary in order to ref