Ye Wang, Yaxiong Wang, Guoshuai Zhao, Xueming Qian
{"title":"知识适应网络促进少数人的课堂强化学习","authors":"Ye Wang, Yaxiong Wang, Guoshuai Zhao, Xueming Qian","doi":"arxiv-2409.11770","DOIUrl":null,"url":null,"abstract":"Few-shot class-incremental learning (FSCIL) aims to incrementally recognize\nnew classes using a few samples while maintaining the performance on previously\nlearned classes. One of the effective methods to solve this challenge is to\nconstruct prototypical evolution classifiers. Despite the advancement achieved\nby most existing methods, the classifier weights are simply initialized using\nmean features. Because representations for new classes are weak and biased, we\nargue such a strategy is suboptimal. In this paper, we tackle this issue from\ntwo aspects. Firstly, thanks to the development of foundation models, we employ\na foundation model, the CLIP, as the network pedestal to provide a general\nrepresentation for each class. Secondly, to generate a more reliable and\ncomprehensive instance representation, we propose a Knowledge Adapter (KA)\nmodule that summarizes the data-specific knowledge from training data and fuses\nit into the general representation. Additionally, to tune the knowledge learned\nfrom the base classes to the upcoming classes, we propose a mechanism of\nIncremental Pseudo Episode Learning (IPEL) by simulating the actual FSCIL.\nTaken together, our proposed method, dubbed as Knowledge Adaptation Network\n(KANet), achieves competitive performance on a wide range of datasets,\nincluding CIFAR100, CUB200, and ImageNet-R.","PeriodicalId":501130,"journal":{"name":"arXiv - CS - Computer Vision and Pattern Recognition","volume":"14 1","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2024-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Knowledge Adaptation Network for Few-Shot Class-Incremental Learning\",\"authors\":\"Ye Wang, Yaxiong Wang, Guoshuai Zhao, Xueming Qian\",\"doi\":\"arxiv-2409.11770\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Few-shot class-incremental learning (FSCIL) aims to incrementally recognize\\nnew classes using a few samples while maintaining the performance on previously\\nlearned classes. One of the effective methods to solve this challenge is to\\nconstruct prototypical evolution classifiers. Despite the advancement achieved\\nby most existing methods, the classifier weights are simply initialized using\\nmean features. Because representations for new classes are weak and biased, we\\nargue such a strategy is suboptimal. In this paper, we tackle this issue from\\ntwo aspects. Firstly, thanks to the development of foundation models, we employ\\na foundation model, the CLIP, as the network pedestal to provide a general\\nrepresentation for each class. Secondly, to generate a more reliable and\\ncomprehensive instance representation, we propose a Knowledge Adapter (KA)\\nmodule that summarizes the data-specific knowledge from training data and fuses\\nit into the general representation. Additionally, to tune the knowledge learned\\nfrom the base classes to the upcoming classes, we propose a mechanism of\\nIncremental Pseudo Episode Learning (IPEL) by simulating the actual FSCIL.\\nTaken together, our proposed method, dubbed as Knowledge Adaptation Network\\n(KANet), achieves competitive performance on a wide range of datasets,\\nincluding CIFAR100, CUB200, and ImageNet-R.\",\"PeriodicalId\":501130,\"journal\":{\"name\":\"arXiv - CS - Computer Vision and Pattern Recognition\",\"volume\":\"14 1\",\"pages\":\"\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2024-09-18\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"arXiv - CS - Computer Vision and Pattern Recognition\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/arxiv-2409.11770\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"arXiv - CS - Computer Vision and Pattern Recognition","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/arxiv-2409.11770","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Knowledge Adaptation Network for Few-Shot Class-Incremental Learning
Few-shot class-incremental learning (FSCIL) aims to incrementally recognize
new classes using a few samples while maintaining the performance on previously
learned classes. One of the effective methods to solve this challenge is to
construct prototypical evolution classifiers. Despite the advancement achieved
by most existing methods, the classifier weights are simply initialized using
mean features. Because representations for new classes are weak and biased, we
argue such a strategy is suboptimal. In this paper, we tackle this issue from
two aspects. Firstly, thanks to the development of foundation models, we employ
a foundation model, the CLIP, as the network pedestal to provide a general
representation for each class. Secondly, to generate a more reliable and
comprehensive instance representation, we propose a Knowledge Adapter (KA)
module that summarizes the data-specific knowledge from training data and fuses
it into the general representation. Additionally, to tune the knowledge learned
from the base classes to the upcoming classes, we propose a mechanism of
Incremental Pseudo Episode Learning (IPEL) by simulating the actual FSCIL.
Taken together, our proposed method, dubbed as Knowledge Adaptation Network
(KANet), achieves competitive performance on a wide range of datasets,
including CIFAR100, CUB200, and ImageNet-R.