{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/cross-to-merge-training-with-class-balance","title":"Cross-to-merge training with class balance strategy for learning with noisy labels","arxiv_id":null,"date":"2024-04-01","proceeding":"Expert Systems with Applications 2024 4","authors":["Qian Zhang","Yi Zhu","Ming Yang","Ge Jin","YingWen Zhu","Qiu Chen"],"abstract":"The collection of large-scale datasets inevitably introduces noisy labels, leading to a substantial degradation in \r\nthe performance of deep neural networks (DNNs). Although sample selection is a mainstream method in the field \r\nof learning with noisy labels, which aims to mitigate the impact of noisy labels during model training, the testing \r\nperformance of these methods exhibits significant fluctuations across different noise rates and types. In this \r\npaper, we propose Cross-to-Merge Training (C2MT), a novel framework that is insensitive to the prior infor\u0002mation in sample selection progress, enhancing model robustness. In practical implementation, using cross\u0002divided training data, two different networks are cross-trained with the co-teaching strategy for several local \r\nrounds, subsequently merged into a unified model by performing federated averages on the parameters of two \r\nmodels periodically. Additionally, we introduce a new class balance strategy, named Median Balance Strategy \r\n(MBS), during the cross-dividing process, which evenly divides the training data into a labeled subset and an \r\nunlabeled subset based on the estimated loss distribution characteristics. Extensive experimental results on both \r\nsynthetic and real-world datasets demonstrate the effectiveness of C2MT. The Code will be available at: https://g \r\nithub.com/LanXiaoPang613/C2MT.","url_abs":"https://doi.org/10.1016/j.eswa.2024.123846","url_pdf":"https://doi.org/10.1016/j.eswa.2024.123846","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"cross-to-merge-training-with-class-balance","repo_url":"https://github.com/LanXiaoPang613/C2MT","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"learning-with-noisy-labels","task_name":"Learning with noisy labels"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/image-classification-on-clothing1m","task":"Image Classification","dataset":"Clothing1M","model":"C2MT","rank_in_archive_order":16,"of":51,"metrics":{"Accuracy":"74.61%"},"uses_additional_data":false},{"leaderboard":"/sota/learning-with-noisy-labels-on-animal","task":"Learning with noisy labels","dataset":"ANIMAL","model":"C2MT","rank_in_archive_order":10,"of":19,"metrics":{"Accuracy":"85.9","ImageNet Pretrained":"NO","Network":"Vgg-19-BN"},"uses_additional_data":false},{"leaderboard":"/sota/learning-with-noisy-labels-on-clothing1m","task":"Learning with noisy labels","dataset":"Clothing1M","model":"C2MT","rank_in_archive_order":3,"of":5,"metrics":{"Test Accuracy":"74.61"},"uses_additional_data":false}],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}