{"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/dimensionality-driven-learning-with-noisy","title":"Dimensionality-Driven Learning with Noisy Labels","arxiv_id":"1806.02612","date":"2018-06-07","proceeding":"ICML 2018 7","authors":["Xingjun Ma","Yisen Wang","Michael E. Houle","Shuo Zhou","Sarah M. Erfani","Shu-Tao Xia","Sudanthi Wijewickrema","James Bailey"],"abstract":"Datasets with significant proportions of noisy (incorrect) class labels\npresent challenges for training accurate Deep Neural Networks (DNNs). We\npropose a new perspective for understanding DNN generalization for such\ndatasets, by investigating the dimensionality of the deep representation\nsubspace of training samples. We show that from a dimensionality perspective,\nDNNs exhibit quite distinctive learning styles when trained with clean labels\nversus when trained with a proportion of noisy labels. Based on this finding,\nwe develop a new dimensionality-driven learning strategy, which monitors the\ndimensionality of subspaces during training and adapts the loss function\naccordingly. We empirically demonstrate that our approach is highly tolerant to\nsignificant proportions of noisy labels, and can effectively learn\nlow-dimensional local subspaces that capture the data distribution.","url_abs":"http://arxiv.org/abs/1806.02612v2","url_pdf":"http://arxiv.org/pdf/1806.02612v2.pdf","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":"dimensionality-driven-learning-with-noisy","repo_url":"https://github.com/xingjunm/dimensionality-driven-learning","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":{"status":"ok"}},{"paper_slug":"dimensionality-driven-learning-with-noisy","repo_url":"https://github.com/ansuini/IntrinsicDimDeep","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"dimensionality-driven-learning-with-noisy","repo_url":"https://github.com/MindSpore-scientific-2/code-3/tree/main/Three-Dimensional-Lip-Motion-Network-for-Text-Independent-Speaker-Recognition-master","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","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":"D2L","rank_in_archive_order":50,"of":51,"metrics":{"Accuracy":"69.47%"},"uses_additional_data":false},{"leaderboard":"/sota/image-classification-on-mini-webvision-1-0","task":"Image Classification","dataset":"mini WebVision 1.0","model":"D2L (Inception-ResNet-v2)","rank_in_archive_order":41,"of":47,"metrics":{"ImageNet Top-1 Accuracy":"57.80","ImageNet Top-5 Accuracy":"81.36","Top-1 Accuracy":"62.68","Top-5 Accuracy":"84.00"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1806.02612","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}