{"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/large-margin-softmax-loss-for-convolutional","title":"Large-Margin Softmax Loss for Convolutional Neural Networks","arxiv_id":"1612.02295","date":"2016-12-07","proceeding":null,"authors":["Weiyang Liu","Yandong Wen","Zhiding Yu","Meng Yang"],"abstract":"Cross-entropy loss together with softmax is arguably one of the most common\nused supervision components in convolutional neural networks (CNNs). Despite\nits simplicity, popularity and excellent performance, the component does not\nexplicitly encourage discriminative learning of features. In this paper, we\npropose a generalized large-margin softmax (L-Softmax) loss which explicitly\nencourages intra-class compactness and inter-class separability between learned\nfeatures. Moreover, L-Softmax not only can adjust the desired margin but also\ncan avoid overfitting. We also show that the L-Softmax loss can be optimized by\ntypical stochastic gradient descent. Extensive experiments on four benchmark\ndatasets demonstrate that the deeply-learned features with L-softmax loss\nbecome more discriminative, hence significantly boosting the performance on a\nvariety of visual classification and verification tasks.","url_abs":"http://arxiv.org/abs/1612.02295v4","url_pdf":"http://arxiv.org/pdf/1612.02295v4.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":"large-margin-softmax-loss-for-convolutional","repo_url":"https://github.com/wy1iu/LargeMargin_Softmax_Loss","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"large-margin-softmax-loss-for-convolutional","repo_url":"https://github.com/amirhfarzaneh/lsoftmax-pytorch","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"classification","task_name":"General Classification"}],"methods":[{"method_slug":"softmax","method_name":"Softmax"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1612.02295","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}