文献类型：期刊文献

英文题名：BFCP: Pursue Better Forward Compatibility Pretraining for Few-Shot Class-Incremental Learning

作者：Fu, Zhiling[1,2];Wang, Zhe[1,2];Xu, Xinlei[1,2];Guo, Wei[1,2];Chi, Ziqiu[1,2];Yang, Hai[2];Du, Wenli[3]

机构：[1]Minist Educ, Key Lab Smart Mfg Energy Chem Proc, Shanghai 200237, Peoples R China;[2]East China Univ Sci & Technol, Dept Comp Sci & Engn, Shanghai 200237, Peoples R China;[3]Minist Educ, Key Lab Smart Mfg Energy Chem Proc, Shanghai 200237, Peoples R China

年份：2025

卷号：36

期号：8

起止页码：14975

外文期刊名：IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS

收录：;EI(收录号：20251218087911);WOS:【SCI-EXPANDED(收录号:WOS:001470700500001)】；

基金：This work was supported in part by the Natural Science Foundation of China under Grant 62476087, in part by Shanghai Municipal Education Commission's Initiative on Artificial Intelligence-Driven Reform of Scientific Research Paradigms and Empowerment of Discipline Leapfrogging, and in part by the National Key Research and Development Program of China under Grant 2022YFB3203500.

语种：英文

外文关键词：Prototypes; Power capacitors; Training; Feature extraction; Incremental learning; Data models; Computational modeling; Accuracy; Knowledge engineering; Smart manufacturing; Few-shot class-incremental learning (FSCIL); forward compatibility; pretraining; prototype learning

摘要：Few-shot class-incremental learning (FSCIL) requires learning new knowledge without forgetting old knowledge. Forward compatibility can reserve space for novel classes while maintaining base class knowledge in incremental learning. Better forward compatibility is crucial for effectively mastering all knowledge, especially when dealing with a few unknown new classes. In this article, we propose the better forward compatibility pretraining (BFCP) to further enhance forward compatibility in FSCIL. We adopt a two-stage training for the backbone network in the base session. First, we train the backbone network at the image-level to enhance its feature extraction capability, enabling the model to extract valuable information from unknown class images. Second, we fine-tune the backbone network at the feature-level with fake prototypes and instances to achieve clustering base classes and reserve space for unknown new classes. For all incremental new sessions, we freeze the backbone network and employ prototype rectification without further training to refine the prototypes of the novel classes. We conduct extensive experiments with different input scales, including federated cross-domain pretraining and cross-domain class-incremental experiments. BFCP efficiently handles both novel and base classes of each incremental session and significantly outperforms state-of-the-art methods, achieving an average accuracy of 63.47% on the CIFAR100 dataset.