FGCM: Noisy Label Learning via Fine-Grained Confidence Modeling
A small portion of mislabeled data can easily limit the performance of deep neural networks (DNNs) due to their high capacity for memorizing random labels. Thus, robust learning from noisy labels has become a key challenge for deep learning due to inadequate datasets with high-quality annotations. M...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-11-01
|
| Series: | Applied Sciences |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2076-3417/12/22/11406 |
| _version_ | 1827645069095600128 |
|---|---|
| author | Shaotian Yan Xiang Tian Rongxin Jiang Yaowu Chen |
| author_facet | Shaotian Yan Xiang Tian Rongxin Jiang Yaowu Chen |
| author_sort | Shaotian Yan |
| collection | DOAJ |
| description | A small portion of mislabeled data can easily limit the performance of deep neural networks (DNNs) due to their high capacity for memorizing random labels. Thus, robust learning from noisy labels has become a key challenge for deep learning due to inadequate datasets with high-quality annotations. Most existing methods involve training models on clean sets by dividing clean samples from noisy ones, resulting in large amounts of mislabeled data being unused. To address this problem, we propose categorizing training samples into five fine-grained clusters based on the difficulty experienced by DNN models when learning them and label correctness. A novel fine-grained confidence modeling (FGCM) framework is proposed to cluster samples into these five categories; with each cluster, FGCM decides whether to accept the cluster data as they are, accept them with label correction, or accept them as unlabeled data. By applying different strategies to the fine-grained clusters, FGCM can better exploit training data than previous methods. Extensive experiments on widely used benchmarks CIFAR-10, CIFAR-100, clothing1M, and WebVision with different ratios and types of label noise demonstrate the superiority of our FGCM. |
| first_indexed | 2024-03-09T18:30:18Z |
| format | Article |
| id | doaj.art-6833377b9e98487e976b4ea99a33a854 |
| institution | Directory Open Access Journal |
| issn | 2076-3417 |
| language | English |
| last_indexed | 2024-03-09T18:30:18Z |
| publishDate | 2022-11-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Applied Sciences |
| spelling | doaj.art-6833377b9e98487e976b4ea99a33a8542023-11-24T07:34:45ZengMDPI AGApplied Sciences2076-34172022-11-0112221140610.3390/app122211406FGCM: Noisy Label Learning via Fine-Grained Confidence ModelingShaotian Yan0Xiang Tian1Rongxin Jiang2Yaowu Chen3College of Biomedical Engineering and Instrument Science, Zhejiang University, Hangzhou 310007, ChinaZhejiang Provincial Key Laboratory for Network Multimedia Technologies, Hangzhou 310007, ChinaZhejiang Provincial Key Laboratory for Network Multimedia Technologies, Hangzhou 310007, ChinaZhejiang University Embedded System Engineering Research Center, Ministry of Education of China, Hangzhou 310007, ChinaA small portion of mislabeled data can easily limit the performance of deep neural networks (DNNs) due to their high capacity for memorizing random labels. Thus, robust learning from noisy labels has become a key challenge for deep learning due to inadequate datasets with high-quality annotations. Most existing methods involve training models on clean sets by dividing clean samples from noisy ones, resulting in large amounts of mislabeled data being unused. To address this problem, we propose categorizing training samples into five fine-grained clusters based on the difficulty experienced by DNN models when learning them and label correctness. A novel fine-grained confidence modeling (FGCM) framework is proposed to cluster samples into these five categories; with each cluster, FGCM decides whether to accept the cluster data as they are, accept them with label correction, or accept them as unlabeled data. By applying different strategies to the fine-grained clusters, FGCM can better exploit training data than previous methods. Extensive experiments on widely used benchmarks CIFAR-10, CIFAR-100, clothing1M, and WebVision with different ratios and types of label noise demonstrate the superiority of our FGCM.https://www.mdpi.com/2076-3417/12/22/11406noisy labeled datarobust learningimage classification |
| spellingShingle | Shaotian Yan Xiang Tian Rongxin Jiang Yaowu Chen FGCM: Noisy Label Learning via Fine-Grained Confidence Modeling Applied Sciences noisy labeled data robust learning image classification |
| title | FGCM: Noisy Label Learning via Fine-Grained Confidence Modeling |
| title_full | FGCM: Noisy Label Learning via Fine-Grained Confidence Modeling |
| title_fullStr | FGCM: Noisy Label Learning via Fine-Grained Confidence Modeling |
| title_full_unstemmed | FGCM: Noisy Label Learning via Fine-Grained Confidence Modeling |
| title_short | FGCM: Noisy Label Learning via Fine-Grained Confidence Modeling |
| title_sort | fgcm noisy label learning via fine grained confidence modeling |
| topic | noisy labeled data robust learning image classification |
| url | https://www.mdpi.com/2076-3417/12/22/11406 |
| work_keys_str_mv | AT shaotianyan fgcmnoisylabellearningviafinegrainedconfidencemodeling AT xiangtian fgcmnoisylabellearningviafinegrainedconfidencemodeling AT rongxinjiang fgcmnoisylabellearningviafinegrainedconfidencemodeling AT yaowuchen fgcmnoisylabellearningviafinegrainedconfidencemodeling |