文献类型：期刊文献

英文题名：Knowledge aggregation networks for class incremental learning

作者：Fu, Zhiling[1,2];Wang, Zhe[1,2];Xu, Xinlei[1,2];Li, Dongdong[2];Yang, Hai[2]

机构：[1]East China Univ Sci & Technol, Key Lab Smart Mfg Energy Chem Proc, Minist Educ, Shanghai 200237, Peoples R China;[2]East China Univ Sci & Technol, Dept Comp Sci & Engn, Shanghai 200237, Peoples R China

年份：2023

卷号：137

外文期刊名：PATTERN RECOGNITION

收录：;EI(收录号：20230413415415);WOS:【SCI-EXPANDED(收录号:WOS:001030278200001)】；

基金：This work is supported by Shanghai Science and Technology Program "Federated based cross-domain and cross -task incremental learning" under Grant No. 2151110 080 0, Natural Science Foundation of China under Grant No. 62076094 , Shanghai Science and Technology Program "Distributed and generative few-shot algorithm and theory research" under Grant No. 2051110 060 0, Chinese Defense Program of Science and Technology under Grant No. 2021- JCJQ-JJ-0041 , China Aerospace Science and Technology Corporation Industry-University-Research Cooperation Foundation of the Eighth Research Institute under Grant No. SAST2021-007.

语种：英文

外文关键词：Class incremental learning; Catastrophic forgetting; Dual-branch network; Knowledge aggregation; Model compression

摘要：Most existing class incremental learning methods rely on storing old exemplars to avoid catastrophic forgetting. However, these methods inevitably face the gradient conflict problem, the inherent conflict between new streaming knowledge and existing knowledge in the gradient direction. To alleviate gradient conflict, this paper reuses the previous knowledge and expands the branch to accommodate new concepts instead of fine-tuning the original models. Specifically, this paper designs a novel dual-branch network called Knowledge Aggregation Networks. The previously trained model is frozen as a branch to retain existing knowledge, and a consistent trainable network is constructed as the other branch to learn new concepts. An adaptive feature fusion module is adopted to dynamically balance the two branches' information during training. Moreover, a model compression stage maintains the dual-branch structure. Extensive experiments on CIFAR-10 0, ImageNet-Sub, and ImageNet show that our method significantly outperforms the other methods and effectively balances stability and plasticity. & COPY; 2023 Elsevier Ltd. All rights reserved.