Open-World Class Discovery with Kernel Networks

TL;DR

This paper introduces CD-KNet-Exp, a deep learning framework that effectively discovers new classes in an open-world setting by transferring and integrating knowledge from old classes using kernel methods.

Contribution

The paper proposes a novel kernel network architecture with an expansion mechanism that systematically combines supervised and unsupervised learning for class discovery.

Findings

Superior performance on benchmark datasets

Effective knowledge transfer from old to new classes

Robust discovery in real-world RF fingerprinting data

Abstract

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.