{"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/making-deep-neural-networks-robust-to-label","title":"Making Deep Neural Networks Robust to Label Noise: a Loss Correction Approach","arxiv_id":"1609.03683","date":"2016-09-13","proceeding":"CVPR 2017 7","authors":["Giorgio Patrini","Alessandro Rozza","Aditya Menon","Richard Nock","Lizhen Qu"],"abstract":"We present a theoretically grounded approach to train deep neural networks,\nincluding recurrent networks, subject to class-dependent label noise. We\npropose two procedures for loss correction that are agnostic to both\napplication domain and network architecture. They simply amount to at most a\nmatrix inversion and multiplication, provided that we know the probability of\neach class being corrupted into another. We further show how one can estimate\nthese probabilities, adapting a recent technique for noise estimation to the\nmulti-class setting, and thus providing an end-to-end framework. Extensive\nexperiments on MNIST, IMDB, CIFAR-10, CIFAR-100 and a large scale dataset of\nclothing images employing a diversity of architectures --- stacking dense,\nconvolutional, pooling, dropout, batch normalization, word embedding, LSTM and\nresidual layers --- demonstrate the noise robustness of our proposals.\nIncidentally, we also prove that, when ReLU is the only non-linearity, the loss\ncurvature is immune to class-dependent label noise.","url_abs":"http://arxiv.org/abs/1609.03683v2","url_pdf":"http://arxiv.org/pdf/1609.03683v2.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":"making-deep-neural-networks-robust-to-label","repo_url":"https://github.com/giorgiop/loss-correction","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null},{"paper_slug":"making-deep-neural-networks-robust-to-label","repo_url":"https://github.com/yikun2019/PENCIL","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"diversity","task_name":"Diversity"},{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"learning-with-noisy-labels","task_name":"Learning with noisy labels"},{"task_slug":"noise-estimation","task_name":"Noise Estimation"}],"methods":[{"method_slug":"lstm","method_name":"LSTM"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"sigmoid-activation","method_name":"Sigmoid Activation"},{"method_slug":"tanh-activation","method_name":"Tanh Activation"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/image-classification-on-clothing1m-using","task":"Image Classification","dataset":"Clothing1M (using clean data)","model":"Forward","rank_in_archive_order":2,"of":9,"metrics":{"Accuracy":"80.27"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-mini-webvision-1-0","task":"Image Classification","dataset":"mini WebVision 1.0","model":"F-Correction (Inception-ResNet-v2)","rank_in_archive_order":42,"of":47,"metrics":{"ImageNet Top-1 Accuracy":"57.36","ImageNet Top-5 Accuracy":"82.36","Top-1 Accuracy":"61.12","Top-5 Accuracy":"82.68"},"uses_additional_data":false},{"leaderboard":"/sota/learning-with-noisy-labels-on-cifar-100n","task":"Learning with noisy labels","dataset":"CIFAR-100N","model":"Backward-T","rank_in_archive_order":17,"of":24,"metrics":{"Accuracy (mean)":"57.14"},"uses_additional_data":false},{"leaderboard":"/sota/learning-with-noisy-labels-on-cifar-100n","task":"Learning with noisy labels","dataset":"CIFAR-100N","model":"Forward-T","rank_in_archive_order":19,"of":24,"metrics":{"Accuracy (mean)":"57.01"},"uses_additional_data":false},{"leaderboard":"/sota/learning-with-noisy-labels-on-cifar-10n","task":"Learning with noisy labels","dataset":"CIFAR-10N-Aggregate","model":"Forward-T","rank_in_archive_order":23,"of":26,"metrics":{"Accuracy (mean)":"88.24"},"uses_additional_data":false},{"leaderboard":"/sota/learning-with-noisy-labels-on-cifar-10n","task":"Learning with noisy labels","dataset":"CIFAR-10N-Aggregate","model":"Backward-T","rank_in_archive_order":24,"of":26,"metrics":{"Accuracy (mean)":"88.13"},"uses_additional_data":false},{"leaderboard":"/sota/learning-with-noisy-labels-on-cifar-10n-1","task":"Learning with noisy labels","dataset":"CIFAR-10N-Random1","model":"Backward-T","rank_in_archive_order":23,"of":24,"metrics":{"Accuracy (mean)":"87.14"},"uses_additional_data":false},{"leaderboard":"/sota/learning-with-noisy-labels-on-cifar-10n-1","task":"Learning with noisy labels","dataset":"CIFAR-10N-Random1","model":"Forward-T","rank_in_archive_order":24,"of":24,"metrics":{"Accuracy (mean)":"86.88"},"uses_additional_data":false},{"leaderboard":"/sota/learning-with-noisy-labels-on-cifar-10n-2","task":"Learning with noisy labels","dataset":"CIFAR-10N-Random2","model":"Backward-T","rank_in_archive_order":22,"of":23,"metrics":{"Accuracy (mean)":"86.28"},"uses_additional_data":false},{"leaderboard":"/sota/learning-with-noisy-labels-on-cifar-10n-2","task":"Learning with noisy labels","dataset":"CIFAR-10N-Random2","model":"Forward-T","rank_in_archive_order":23,"of":23,"metrics":{"Accuracy (mean)":"86.14"},"uses_additional_data":false},{"leaderboard":"/sota/learning-with-noisy-labels-on-cifar-10n-3","task":"Learning with noisy labels","dataset":"CIFAR-10N-Random3","model":"Forward-T","rank_in_archive_order":21,"of":23,"metrics":{"Accuracy (mean)":"87.04"},"uses_additional_data":false},{"leaderboard":"/sota/learning-with-noisy-labels-on-cifar-10n-3","task":"Learning with noisy labels","dataset":"CIFAR-10N-Random3","model":"Backward-T","rank_in_archive_order":22,"of":23,"metrics":{"Accuracy (mean)":"86.86"},"uses_additional_data":false},{"leaderboard":"/sota/learning-with-noisy-labels-on-cifar-10n-worst","task":"Learning with noisy labels","dataset":"CIFAR-10N-Worst","model":"Forward-T","rank_in_archive_order":23,"of":25,"metrics":{"Accuracy (mean)":"79.79"},"uses_additional_data":false},{"leaderboard":"/sota/learning-with-noisy-labels-on-cifar-10n-worst","task":"Learning with noisy labels","dataset":"CIFAR-10N-Worst","model":"Backward-T","rank_in_archive_order":25,"of":25,"metrics":{"Accuracy (mean)":"77.61"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1609.03683","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"1609.03683"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-24T18:15:14+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/yikun2019/PENCIL","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/giorgiop/loss-correction","reach":null}],"summary":{"ran_draft_wrong":2},"by_repo_kind":{"official":{"samples":1,"ran":1,"repositories":1},"listed":{"samples":1,"ran":1,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":2,"samples":[{"code_sha256_prefix":"adf3acfa302cc3ef","entry":"build_file_name","repo":"giorgiop/loss-correction","repo_kind":"official","path":"experiment.py","file_url":"https://github.com/giorgiop/loss-correction/blob/HEAD/experiment.py","link_basis":"first_harvest_node","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"adf3acfa302cc3ef"}},{"code_sha256_prefix":"1f0bb699e23ab672","entry":"unpickle","repo":"yikun2019/PENCIL","repo_kind":"listed","path":"addnoise_AN.py","file_url":"https://github.com/yikun2019/PENCIL/blob/HEAD/addnoise_AN.py","link_basis":"first_harvest_node","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"1f0bb699e23ab672"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}