{"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/negative-margin-matters-understanding-margin","title":"Negative Margin Matters: Understanding Margin in Few-shot Classification","arxiv_id":"2003.12060","date":"2020-03-26","proceeding":"ECCV 2020 8","authors":["Bin Liu","Yue Cao","Yutong Lin","Qi Li","Zheng Zhang","Mingsheng Long","Han Hu"],"abstract":"This paper introduces a negative margin loss to metric learning based few-shot learning methods. The negative margin loss significantly outperforms regular softmax loss, and achieves state-of-the-art accuracy on three standard few-shot classification benchmarks with few bells and whistles. These results are contrary to the common practice in the metric learning field, that the margin is zero or positive. To understand why the negative margin loss performs well for the few-shot classification, we analyze the discriminability of learned features w.r.t different margins for training and novel classes, both empirically and theoretically. We find that although negative margin reduces the feature discriminability for training classes, it may also avoid falsely mapping samples of the same novel class to multiple peaks or clusters, and thus benefit the discrimination of novel classes. Code is available at https://github.com/bl0/negative-margin.few-shot.","url_abs":"https://arxiv.org/abs/2003.12060v1","url_pdf":"https://arxiv.org/pdf/2003.12060v1.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":"negative-margin-matters-understanding-margin","repo_url":"https://github.com/bl0/negative-margin.few-shot","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"NOASSERTION"}}],"tasks":[{"task_slug":"classification-1","task_name":"Classification"},{"task_slug":"few-shot-image-classification","task_name":"Few-Shot Image Classification"},{"task_slug":"few-shot-learning","task_name":"Few-Shot Learning"},{"task_slug":"classification","task_name":"General Classification"},{"task_slug":"metric-learning","task_name":"Metric Learning"}],"methods":[{"method_slug":"softmax","method_name":"Softmax"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/few-shot-image-classification-on-cub-200-5-1","task":"Few-Shot Image Classification","dataset":"CUB 200 5-way 1-shot","model":"Neg-Margin","rank_in_archive_order":26,"of":36,"metrics":{"Accuracy":"72.66"},"uses_additional_data":false},{"leaderboard":"/sota/few-shot-image-classification-on-cub-200-5","task":"Few-Shot Image Classification","dataset":"CUB 200 5-way 5-shot","model":"Neg-Margin","rank_in_archive_order":20,"of":32,"metrics":{"Accuracy":"89.40"},"uses_additional_data":false},{"leaderboard":"/sota/few-shot-image-classification-on-mini-1","task":"Few-Shot Image Classification","dataset":"Mini-ImageNet - 1-Shot Learning","model":"Neg-Margin","rank_in_archive_order":12,"of":16,"metrics":{"Accuracy":"63.85"},"uses_additional_data":false},{"leaderboard":"/sota/few-shot-image-classification-on-mini-10","task":"Few-Shot Image Classification","dataset":"Mini-ImageNet to CUB - 5 shot learning","model":"Neg-Margin","rank_in_archive_order":2,"of":2,"metrics":{"Accuracy":"69.30"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2003.12060","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}