Class Incremental Learning With Less Forgetting Direction and Equilibrium Point
- 10 October 2024
- journal article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems for Video Technology
- Vol. 35 (2) , 1150-1164
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 problem, but seldom from the aspect of parameter regularization. However, existing parameter regularization methods struggle to achieve an appropriate balance between old and new tasks. To bring it back to CIL, we first propose constrained incremental learning with less forgetting direction (LFD) to leave more plasticity for the new task under a strong stability constraint for old tasks. Specifically, the new parameters are constrained to be close to the LFD of old tasks instead of a single group of old parameters. To validate the effectiveness of this regularization, we investigate the connectivity between the old parameters and the new parameters, and additionally find that a higher accuracy interval exists along the linear connection. Therefore, we further propose a post-processing procedure to find an equilibrium point in this interval for better balance between old and new tasks. Extensive classification experiments on CIFAR-100, ImageNet-100, and ImageNet-1K show our method can significantly improve performance compared with existing CIL methods and the object detection experiments on PASCAL-VOC show its broad generality on other tasks.Keywords
Funding Information
- National Science and Technology Major Project (2021ZD0112001)
- National Natural Science Foundation of China (U23A20286)
- China Postdoctoral Science Foundation (2023TQ0046, 2023M740529)
- Postdoctoral Fellowship Program (Grade B) of China Postdoctoral Science Foundation (GZB20240120)
This publication has 57 references indexed in Scilit:
- Compositional Prompting for Anti-Forgetting in Domain Incremental LearningInternational Journal of Computer Vision, 2024
- Balancing Between Forgetting and Acquisition in Incremental Subpopulation Learning2022
- FOSTER: Feature Boosting and Compression for Class-Incremental Learning2022
- Face.evoLVe: A cross-platform library for high-performance face analyticsNeurocomputing, 2022
- PODNet: Pooled Outputs Distillation for Small-Tasks Incremental Learning2020
- Continual learning of context-dependent processing in neural networksNature Machine Intelligence, 2019
- Memory Aware Synapses: Learning What (not) to Forget2018
- Riemannian Walk for Incremental Learning: Understanding Forgetting and Intransigence2018
- The Pascal Visual Object Classes (VOC) ChallengeInternational Journal of Computer Vision, 2009
- Introduction to Smooth Manifolds2003