{"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/improving-zero-shot-learning-by-mitigating","title":"Improving zero-shot learning by mitigating the hubness problem","arxiv_id":"1412.6568","date":"2014-12-20","proceeding":null,"authors":["Georgiana Dinu","Angeliki Lazaridou","Marco Baroni"],"abstract":"The zero-shot paradigm exploits vector-based word representations extracted\nfrom text corpora with unsupervised methods to learn general mapping functions\nfrom other feature spaces onto word space, where the words associated to the\nnearest neighbours of the mapped vectors are used as their linguistic labels.\nWe show that the neighbourhoods of the mapped elements are strongly polluted by\nhubs, vectors that tend to be near a high proportion of items, pushing their\ncorrect labels down the neighbour list. After illustrating the problem\nempirically, we propose a simple method to correct it by taking the proximity\ndistribution of potential neighbours across many mapped vectors into account.\nWe show that this correction leads to consistent improvements in realistic\nzero-shot experiments in the cross-lingual, image labeling and image retrieval\ndomains.","url_abs":"http://arxiv.org/abs/1412.6568v3","url_pdf":"http://arxiv.org/pdf/1412.6568v3.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":"improving-zero-shot-learning-by-mitigating","repo_url":"https://github.com/Babylonpartners/fastText_multilingual","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"BSD-3-Clause"}},{"paper_slug":"improving-zero-shot-learning-by-mitigating","repo_url":"https://github.com/babylonhealth/fastText_multilingual","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"BSD-3-Clause"}},{"paper_slug":"improving-zero-shot-learning-by-mitigating","repo_url":"https://github.com/facebookresearch/MUSE","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"improving-zero-shot-learning-by-mitigating","repo_url":"https://github.com/jiajunhua/facebookresearch-MUSE","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"improving-zero-shot-learning-by-mitigating","repo_url":"https://github.com/ssharoff/cognates","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"GPL-3.0"}}],"tasks":[{"task_slug":"image-retrieval","task_name":"Image Retrieval"},{"task_slug":"retrieval","task_name":"Retrieval"},{"task_slug":"zero-shot-learning","task_name":"Zero-Shot Learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1412.6568","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}