Class Incremental Learning from First Principles: A Review
Abstract: Continual learning systems attempt to efficiently learn over time without forgetting previously acquired knowledge. In recent years, there has been an explosion of work on continual learning, mainly focused on the class-incremental learning (CIL) setting. In this review, we take a step back and reconsider the CIL problem. We reexamine the problem definition and describe its unique challenges, contextualize existing solutions by analyzing non-continual approaches, and investigate the implications of various problem configurations. Our goal is to provide an alternative perspective to existing work on CIL and direct attention toward unexplored aspects of the problem.
Submission Length: Long submission (more than 12 pages of main content)
Changes Since Last Submission: We have updated the draft in response to reviewer feedback. Specifically, we have added the following content:
- Section 1: table summarizing previous CL/CIL reviews
- Section 2: discussion of related CL problem statements (FSCIL, TIL, DIL, OCL)
- Section 2: description of a potential real-world application which fits the resource-constrained CIL setting
- Section 5: new subsection discussing pretraining-based and few-shot CIL approaches
- Section 7: further discussion on future work, actionable insights for CL researchers
Assigned Action Editor: ~Erin_Grant1
Submission Number: 3355
Loading