Squeezing More Past Knowledge for Online Class-Incremental Continual Learning
文献类型：期刊论文
| 作者 | Da Yu; Mingyi Zhang ; Mantian Li; Fusheng Zha; Junge Zhang; Lining Sun; Kaiqi Huang
|
| 刊名 | IEEE/CAA Journal of Automatica Sinica
 |
| 出版日期 | 2023 |
| 卷号 | 10期号:3页码:722-736 |
| ISSN号 | 2329-9266 |
| 关键词 | |
| DOI | 10.1109/JAS.2023.123090 |
| 英文摘要 | Continual learning (CL) studies the problem of learning to accumulate knowledge over time from a stream of data. A crucial challenge is that neural networks suffer from performance degradation on previously seen data, known as catastrophic forgetting, due to allowing parameter sharing. In this work, we consider a more practical online class-incremental CL setting, where the model learns new samples in an online manner and may continuously experience new classes. Moreover, prior knowledge is unavailable during training and evaluation. Existing works usually explore sample usages from a single dimension, which ignores a lot of valuable supervisory information. To better tackle the setting, we propose a novel replay-based CL method, which leverages multi-level representations produced by the intermediate process of training samples for replay and strengthens supervision to consolidate previous knowledge. Specifically, besides the previous raw samples, we store the corresponding logits and features in the memory. Furthermore, to imitate the prediction of the past model, we construct extra constraints by leveraging multi-level information stored in the memory. With the same number of samples for replay, our method can use more past knowledge to prevent interference. We conduct extensive evaluations on several popular CL datasets, and experiments show that our method consistently outperforms state-of-the-art methods with various sizes of episodic memory. We further provide a detailed analysis of these results and demonstrate that our method is more viable in practical scenarios. |
| 源URL | [http://ir.ia.ac.cn/handle/173211/51184]  |
| 专题 | 自动化研究所_学术期刊_IEEE/CAA Journal of Automatica Sinica |
| 推荐引用方式 GB/T 7714 | Da Yu,Mingyi Zhang,Mantian Li,et al. Squeezing More Past Knowledge for Online Class-Incremental Continual Learning[J]. IEEE/CAA Journal of Automatica Sinica,2023,10(3):722-736. |
| APA | Da Yu.,Mingyi Zhang.,Mantian Li.,Fusheng Zha.,Junge Zhang.,...&Kaiqi Huang.(2023).Squeezing More Past Knowledge for Online Class-Incremental Continual Learning.IEEE/CAA Journal of Automatica Sinica,10(3),722-736. |
| MLA | Da Yu,et al."Squeezing More Past Knowledge for Online Class-Incremental Continual Learning".IEEE/CAA Journal of Automatica Sinica 10.3(2023):722-736. |
入库方式： OAI收割
来源：自动化研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。