An improved sample selection framework for learning with noisy labels.
Deep neural networks have powerful memory capabilities, yet they frequently suffer from overfitting to noisy labels, leading to a decline in classification and generalization performance. To address this issue, sample selection methods that filter out potentially clean labels have been proposed. How...
Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Public Library of Science (PLoS)
2024-01-01
|
| Series: | PLoS ONE |
| Online Access: | https://doi.org/10.1371/journal.pone.0309841 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1846129412373741568 |
|---|---|
| author | Qian Zhang Yi Zhu Ming Yang Ge Jin Yingwen Zhu Yanjun Lu Yu Zou Qiu Chen |
| author_facet | Qian Zhang Yi Zhu Ming Yang Ge Jin Yingwen Zhu Yanjun Lu Yu Zou Qiu Chen |
| author_sort | Qian Zhang |
| collection | DOAJ |
| description | Deep neural networks have powerful memory capabilities, yet they frequently suffer from overfitting to noisy labels, leading to a decline in classification and generalization performance. To address this issue, sample selection methods that filter out potentially clean labels have been proposed. However, there is a significant gap in size between the filtered, possibly clean subset and the unlabeled subset, which becomes particularly pronounced at high-noise rates. Consequently, this results in underutilizing label-free samples in sample selection methods, leaving room for performance improvement. This study introduces an enhanced sample selection framework with an oversampling strategy (SOS) to overcome this limitation. This framework leverages the valuable information contained in label-free instances to enhance model performance by combining an SOS with state-of-the-art sample selection methods. We validate the effectiveness of SOS through extensive experiments conducted on both synthetic noisy datasets and real-world datasets such as CIFAR, WebVision, and Clothing1M. The source code for SOS will be made available at https://github.com/LanXiaoPang613/SOS. |
| format | Article |
| id | doaj-art-a8ef4eed89764935b909ee9bec2dc5d7 |
| institution | Kabale University |
| issn | 1932-6203 |
| language | English |
| publishDate | 2024-01-01 |
| publisher | Public Library of Science (PLoS) |
| record_format | Article |
| series | PLoS ONE |
| spelling | doaj-art-a8ef4eed89764935b909ee9bec2dc5d72024-12-10T05:32:30ZengPublic Library of Science (PLoS)PLoS ONE1932-62032024-01-011912e030984110.1371/journal.pone.0309841An improved sample selection framework for learning with noisy labels.Qian ZhangYi ZhuMing YangGe JinYingwen ZhuYanjun LuYu ZouQiu ChenDeep neural networks have powerful memory capabilities, yet they frequently suffer from overfitting to noisy labels, leading to a decline in classification and generalization performance. To address this issue, sample selection methods that filter out potentially clean labels have been proposed. However, there is a significant gap in size between the filtered, possibly clean subset and the unlabeled subset, which becomes particularly pronounced at high-noise rates. Consequently, this results in underutilizing label-free samples in sample selection methods, leaving room for performance improvement. This study introduces an enhanced sample selection framework with an oversampling strategy (SOS) to overcome this limitation. This framework leverages the valuable information contained in label-free instances to enhance model performance by combining an SOS with state-of-the-art sample selection methods. We validate the effectiveness of SOS through extensive experiments conducted on both synthetic noisy datasets and real-world datasets such as CIFAR, WebVision, and Clothing1M. The source code for SOS will be made available at https://github.com/LanXiaoPang613/SOS.https://doi.org/10.1371/journal.pone.0309841 |
| spellingShingle | Qian Zhang Yi Zhu Ming Yang Ge Jin Yingwen Zhu Yanjun Lu Yu Zou Qiu Chen An improved sample selection framework for learning with noisy labels. PLoS ONE |
| title | An improved sample selection framework for learning with noisy labels. |
| title_full | An improved sample selection framework for learning with noisy labels. |
| title_fullStr | An improved sample selection framework for learning with noisy labels. |
| title_full_unstemmed | An improved sample selection framework for learning with noisy labels. |
| title_short | An improved sample selection framework for learning with noisy labels. |
| title_sort | improved sample selection framework for learning with noisy labels |
| url | https://doi.org/10.1371/journal.pone.0309841 |
| work_keys_str_mv | AT qianzhang animprovedsampleselectionframeworkforlearningwithnoisylabels AT yizhu animprovedsampleselectionframeworkforlearningwithnoisylabels AT mingyang animprovedsampleselectionframeworkforlearningwithnoisylabels AT gejin animprovedsampleselectionframeworkforlearningwithnoisylabels AT yingwenzhu animprovedsampleselectionframeworkforlearningwithnoisylabels AT yanjunlu animprovedsampleselectionframeworkforlearningwithnoisylabels AT yuzou animprovedsampleselectionframeworkforlearningwithnoisylabels AT qiuchen animprovedsampleselectionframeworkforlearningwithnoisylabels AT qianzhang improvedsampleselectionframeworkforlearningwithnoisylabels AT yizhu improvedsampleselectionframeworkforlearningwithnoisylabels AT mingyang improvedsampleselectionframeworkforlearningwithnoisylabels AT gejin improvedsampleselectionframeworkforlearningwithnoisylabels AT yingwenzhu improvedsampleselectionframeworkforlearningwithnoisylabels AT yanjunlu improvedsampleselectionframeworkforlearningwithnoisylabels AT yuzou improvedsampleselectionframeworkforlearningwithnoisylabels AT qiuchen improvedsampleselectionframeworkforlearningwithnoisylabels |