CASIA OpenIR  > 多模态人工智能系统全国重点实验室
Improving metric-based few-shot learning with dynamically scaled softmax loss
Zhang, Yu1; Zuo, Xin1; Zheng, Xuxu2; Gao, Xiaoyong1; Wang, Bo3,4,7; Hu, Weiming3,5,6
Source PublicationIMAGE AND VISION COMPUTING
ISSN0262-8856
2023-12-01
Volume140Pages:15
Corresponding AuthorWang, Bo(wangbo@ia.ac.cn)
AbstractThe metric-based learning framework has been widely used in data-scarce few-shot visual classification. However, the current loss function limits the effectiveness of metric learning. One issue is that the nearest neighbor classification technique used greatly narrows the value range of similarity between the query and class prototypes, which limits the guiding ability of the loss function. The other issue is that the episode-based training setting randomizes the class combination in each iteration, which reduces the perception of the traditional softmax losses for effective learning from episodes with various data distributions.To solve these problems, we first review some variants of the softmax loss from a unified perspective, and then propose a novel DynamicallyScaled Softmax Loss (DSSL). By adding a probability regulator (for scaling probabilities) and a loss regulator (for scaling losses), the loss function can adaptively adjust the prediction distribution and the training weights of the samples, which forces the model to focus on more informative samples. Finally, we found the proposed DSSL strategy for few-shot classifiers can achieve competitive results on four generic benchmarks and a fine-grained benchmark, demonstrating the effectiveness of improving the distinguishability (for base classes) and generalizability (for novel classes) of the learned feature space.
KeywordFew-shot learning Metric-based learning framework Softmax loss improvement
DOI10.1016/j.imavis.2023.104860
WOS KeywordALIGNMENT
Indexed BySCI
Language英语
Funding ProjectBeijing Natural Science Foundation[62036011] ; Beijing Natural Science Foundation[62192782] ; Beijing Natural Science Foundation[61721004] ; Beijing Natural Science Foundation[U2033210] ; Major Projects of Guangdong Edu-cation Department for Foundation Research and Applied Research[L223003] ; Guangdong Provincial University Innovation Team Project[2017KZDXM081] ; Guangdong Provincial University Innovation Team Project[2018KZDXM066] ; [2020KCXTD045]
Funding OrganizationBeijing Natural Science Foundation ; Major Projects of Guangdong Edu-cation Department for Foundation Research and Applied Research ; Guangdong Provincial University Innovation Team Project
WOS Research AreaComputer Science ; Engineering ; Optics
WOS SubjectComputer Science, Artificial Intelligence ; Computer Science, Software Engineering ; Computer Science, Theory & Methods ; Engineering, Electrical & Electronic ; Optics
WOS IDWOS:001110250500001
PublisherELSEVIER
Citation statistics
Cited Times:1[WOS]   [WOS Record]     [Related Records in WOS]
Document Type期刊论文
Identifierhttp://ir.ia.ac.cn/handle/173211/55108
Collection多模态人工智能系统全国重点实验室
Corresponding AuthorWang, Bo
Affiliation1.China Univ Petr, Beijing 102249, Peoples R China
2.Chinese Acad Sci, Data Intelligence Syst Res Ctr, Inst Comp Technol, Beijing 100190, Peoples R China
3.Chinese Acad Sci, Inst Automat, Natl Lab Pattern Recognit, Beijing 100190, Peoples R China
4.Peking Univ, Sch Software & Microelect, Beijing 100871, Peoples R China
5.Univ Chinese Acad Sci, Sch Artificial Intelligence, Beijing 100190, Peoples R China
6.ShanghaiTech Univ, Sch Informat Sci & Technol, Shanghai 201210, Peoples R China
7.Chinese Acad Sci, Natl Lab Pattern Recognit, Inst Automat, 95 Zhongguancun East Rd, Beijing 100190, Peoples R China
Corresponding Author AffilicationChinese Acad Sci, Inst Automat, Natl Lab Pattern Recognit, Beijing 100190, Peoples R China
Recommended Citation
GB/T 7714
Zhang, Yu,Zuo, Xin,Zheng, Xuxu,et al. Improving metric-based few-shot learning with dynamically scaled softmax loss[J]. IMAGE AND VISION COMPUTING,2023,140:15.
APA Zhang, Yu,Zuo, Xin,Zheng, Xuxu,Gao, Xiaoyong,Wang, Bo,&Hu, Weiming.(2023).Improving metric-based few-shot learning with dynamically scaled softmax loss.IMAGE AND VISION COMPUTING,140,15.
MLA Zhang, Yu,et al."Improving metric-based few-shot learning with dynamically scaled softmax loss".IMAGE AND VISION COMPUTING 140(2023):15.
Files in This Item:
There are no files associated with this item.
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[Zhang, Yu]'s Articles
[Zuo, Xin]'s Articles
[Zheng, Xuxu]'s Articles
Baidu academic
Similar articles in Baidu academic
[Zhang, Yu]'s Articles
[Zuo, Xin]'s Articles
[Zheng, Xuxu]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[Zhang, Yu]'s Articles
[Zuo, Xin]'s Articles
[Zheng, Xuxu]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.