IEEE

Class Incremental Learning With Less Forgetting Direction and Equilibrium Point

Abstract: Catastrophic forgetting is the core problem of class incremental learning (CIL). Existing work mainly adopts memory replay, knowledge distillation, and dynamic architecture to alleviate this ...
IEEE

Semantic Bridging and Feature Anchoring for Class Incremental Learning

Abstract: The purpose of class-incremental learning is to continually assimilate new classes and preserving the knowledge of learned classes. An important issue is that the learned knowledge would be ...