{"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/adversarial-extreme-multi-label","title":"Adversarial Extreme Multi-label Classification","arxiv_id":"1803.01570","date":"2018-03-05","proceeding":null,"authors":["Rohit Babbar","Bernhard Schölkopf"],"abstract":"The goal in extreme multi-label classification is to learn a classifier which\ncan assign a small subset of relevant labels to an instance from an extremely\nlarge set of target labels. Datasets in extreme classification exhibit a long\ntail of labels which have small number of positive training instances. In this\nwork, we pose the learning task in extreme classification with large number of\ntail-labels as learning in the presence of adversarial perturbations. This view\nmotivates a robust optimization framework and equivalence to a corresponding\nregularized objective.\n  Under the proposed robustness framework, we demonstrate efficacy of Hamming\nloss for tail-label detection in extreme classification. The equivalent\nregularized objective, in combination with proximal gradient based\noptimization, performs better than state-of-the-art methods on propensity\nscored versions of precision@k and nDCG@k(upto 20% relative improvement over\nPFastreXML - a leading tree-based approach and 60% relative improvement over\nSLEEC - a leading label-embedding approach). Furthermore, we also highlight the\nsub-optimality of a sparse solver in a widely used package for large-scale\nlinear classification, which is interesting in its own right. We also\ninvestigate the spectral properties of label graphs for providing novel\ninsights towards understanding the conditions governing the performance of\nHamming loss based one-vs-rest scheme vis-\\`a-vis label embedding methods.","url_abs":"http://arxiv.org/abs/1803.01570v1","url_pdf":"http://arxiv.org/pdf/1803.01570v1.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":"adversarial-extreme-multi-label","repo_url":"https://github.com/xmc-aalto/proxml","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"none","reach":{"status":"ok","spdx":"BSD-3-Clause"}}],"tasks":[{"task_slug":"classification-1","task_name":"Classification"},{"task_slug":"extreme-multi-label-classification","task_name":"Extreme Multi-Label Classification"},{"task_slug":"classification","task_name":"General Classification"},{"task_slug":"multi-label-classification-2","task_name":"MUlTI-LABEL-ClASSIFICATION"},{"task_slug":"multi-label-classification","task_name":"Multi-Label Classification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1803.01570","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}