{"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/rethinking-feature-distribution-for-loss","title":"Rethinking Feature Distribution for Loss Functions in Image Classification","arxiv_id":"1803.02988","date":"2018-03-08","proceeding":"CVPR 2018 6","authors":["Weitao Wan","Yuanyi Zhong","Tianpeng Li","Jiansheng Chen"],"abstract":"We propose a large-margin Gaussian Mixture (L-GM) loss for deep neural\nnetworks in classification tasks. Different from the softmax cross-entropy\nloss, our proposal is established on the assumption that the deep features of\nthe training set follow a Gaussian Mixture distribution. By involving a\nclassification margin and a likelihood regularization, the L-GM loss\nfacilitates both a high classification performance and an accurate modeling of\nthe training feature distribution. As such, the L-GM loss is superior to the\nsoftmax loss and its major variants in the sense that besides classification,\nit can be readily used to distinguish abnormal inputs, such as the adversarial\nexamples, based on their features' likelihood to the training feature\ndistribution. Extensive experiments on various recognition benchmarks like\nMNIST, CIFAR, ImageNet and LFW, as well as on adversarial examples demonstrate\nthe effectiveness of our proposal.","url_abs":"http://arxiv.org/abs/1803.02988v1","url_pdf":"http://arxiv.org/pdf/1803.02988v1.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":"rethinking-feature-distribution-for-loss","repo_url":"https://github.com/yuyijie1995/gluon_GMLoss","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"mxnet","reach":{"status":"ok"}}],"tasks":[{"task_slug":"classification-1","task_name":"Classification"},{"task_slug":"classification","task_name":"General Classification"},{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"image-classification","task_name":"image-classification"}],"methods":[{"method_slug":"softmax","method_name":"Softmax"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1803.02988","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}