{"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/confident-multiple-choice-learning","title":"Confident Multiple Choice Learning","arxiv_id":"1706.03475","date":"2017-06-12","proceeding":"ICML 2017 8","authors":["Kimin Lee","Changho Hwang","Kyoungsoo Park","Jinwoo Shin"],"abstract":"Ensemble methods are arguably the most trustworthy techniques for boosting\nthe performance of machine learning models. Popular independent ensembles (IE)\nrelying on naive averaging/voting scheme have been of typical choice for most\napplications involving deep neural networks, but they do not consider advanced\ncollaboration among ensemble models. In this paper, we propose new ensemble\nmethods specialized for deep neural networks, called confident multiple choice\nlearning (CMCL): it is a variant of multiple choice learning (MCL) via\naddressing its overconfidence issue.In particular, the proposed major\ncomponents of CMCL beyond the original MCL scheme are (i) new loss, i.e.,\nconfident oracle loss, (ii) new architecture, i.e., feature sharing and (iii)\nnew training method, i.e., stochastic labeling. We demonstrate the effect of\nCMCL via experiments on the image classification on CIFAR and SVHN, and the\nforeground-background segmentation on the iCoseg. In particular, CMCL using 5\nresidual networks provides 14.05% and 6.60% relative reductions in the top-1\nerror rates from the corresponding IE scheme for the classification task on\nCIFAR and SVHN, respectively.","url_abs":"http://arxiv.org/abs/1706.03475v2","url_pdf":"http://arxiv.org/pdf/1706.03475v2.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":"confident-multiple-choice-learning","repo_url":"https://github.com/chhwang/cmcl","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"confident-multiple-choice-learning","repo_url":"https://github.com/celsolbm/CMCL_Sequence","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"classification","task_name":"General Classification"},{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"multiple-choice","task_name":"Multiple-choice"},{"task_slug":"image-classification","task_name":"image-classification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1706.03475","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}