Open-World Class Discovery with Kernel Networks
This work is significant for researchers and practitioners working on continually discovering new categories in evolving real-world data streams, particularly in scenarios where data is initially partially labeled.
This paper addresses the Open-World Class Discovery problem, where the goal is to discover new classes from unlabeled test samples using knowledge from labeled old classes. The proposed CD-KNet-Exp framework, which bridges supervised and unsupervised information, achieved superior performance on three benchmark datasets and a radio frequency fingerprinting dataset compared to competing methods.
We study an Open-World Class Discovery problem in which, given labeled training samples from old classes, we need to discover new classes from unlabeled test samples. There are two critical challenges to addressing this paradigm: (a) transferring knowledge from old to new classes, and (b) incorporating knowledge learned from new classes back to the original model. We propose Class Discovery Kernel Network with Expansion (CD-KNet-Exp), a deep learning framework, which utilizes the Hilbert Schmidt Independence Criterion to bridge supervised and unsupervised information together in a systematic way, such that the learned knowledge from old classes is distilled appropriately for discovering new classes. Compared to competing methods, CD-KNet-Exp shows superior performance on three publicly available benchmark datasets and a challenging real-world radio frequency fingerprinting dataset.