Dynamic Feature Learning and Matching for Class-Incremental Learning

arXiv:2405.08533v1 Announce Type: new
Abstract: Class-incremental learning (CIL) has emerged as a means to learn new classes incrementally without catastrophic forgetting of previous classes. Recently, CIL has undergone a paradigm shift towards dynamic architectures due to their superior performance. However, these models are still limited by the following aspects: (i) Data augmentation (DA), which are tightly coupled with CIL, remains under-explored in dynamic architecture scenarios. (ii) Feature representation. The discriminativeness of dynamic feature are sub-optimal and possess potential for refinement. …

abstract architectures arxiv augmentation catastrophic forgetting class cs.cv data dynamic feature however incremental incremental learning learn paradigm performance shift type