{"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/deep-cocktail-network-multi-source","title":"Deep Cocktail Network: Multi-source Unsupervised Domain Adaptation with Category Shift","arxiv_id":"1803.00830","date":"2018-03-02","proceeding":"CVPR 2018 6","authors":["Ruijia Xu","Ziliang Chen","WangMeng Zuo","Junjie Yan","Liang Lin"],"abstract":"Unsupervised domain adaptation (UDA) conventionally assumes labeled source\nsamples coming from a single underlying source distribution. Whereas in\npractical scenario, labeled data are typically collected from diverse sources.\nThe multiple sources are different not only from the target but also from each\nother, thus, domain adaptater should not be modeled in the same way. Moreover,\nthose sources may not completely share their categories, which further brings a\nnew transfer challenge called category shift. In this paper, we propose a deep\ncocktail network (DCTN) to battle the domain and category shifts among multiple\nsources. Motivated by the theoretical results in \\cite{mansour2009domain}, the\ntarget distribution can be represented as the weighted combination of source\ndistributions, and, the multi-source unsupervised domain adaptation via DCTN is\nthen performed as two alternating steps: i) It deploys multi-way adversarial\nlearning to minimize the discrepancy between the target and each of the\nmultiple source domains, which also obtains the source-specific perplexity\nscores to denote the possibilities that a target sample belongs to different\nsource domains. ii) The multi-source category classifiers are integrated with\nthe perplexity scores to classify target sample, and the pseudo-labeled target\nsamples together with source samples are utilized to update the multi-source\ncategory classifier and the feature extractor. We evaluate DCTN in three domain\nadaptation benchmarks, which clearly demonstrate the superiority of our\nframework.","url_abs":"http://arxiv.org/abs/1803.00830v1","url_pdf":"http://arxiv.org/pdf/1803.00830v1.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":"deep-cocktail-network-multi-source","repo_url":"https://github.com/HCPLab-SYSU/MSDA","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"domain-adaptation","task_name":"Domain Adaptation"},{"task_slug":"multi-source-unsupervised-domain-adaptation","task_name":"Multi-Source Unsupervised Domain Adaptation"},{"task_slug":"unsupervised-domain-adaptation","task_name":"Unsupervised Domain Adaptation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1803.00830","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}