{"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/manifold-criterion-guided-transfer-learning","title":"Manifold Criterion Guided Transfer Learning via Intermediate Domain Generation","arxiv_id":"1903.10211","date":"2019-03-25","proceeding":null,"authors":["Lei Zhang","Shan-Shan Wang","Guang-Bin Huang","WangMeng Zuo","Jian Yang","David Zhang"],"abstract":"In many practical transfer learning scenarios, the feature distribution is\ndifferent across the source and target domains (i.e. non-i.i.d.). Maximum mean\ndiscrepancy (MMD), as a domain discrepancy metric, has achieved promising\nperformance in unsupervised domain adaptation (DA). We argue that MMD-based DA\nmethods ignore the data locality structure, which, to some extent, would cause\nthe negative transfer effect. The locality plays an important role in\nminimizing the nonlinear local domain discrepancy underlying the marginal\ndistributions. For better exploiting the domain locality, a novel local\ngenerative discrepancy metric (LGDM) based intermediate domain generation\nlearning called Manifold Criterion guided Transfer Learning (MCTL) is proposed\nin this paper. The merits of the proposed MCTL are four-fold: 1) the concept of\nmanifold criterion (MC) is first proposed as a measure validating the\ndistribution matching across domains, and domain adaptation is achieved if the\nMC is satisfied; 2) the proposed MC can well guide the generation of the\nintermediate domain sharing similar distribution with the target domain, by\nminimizing the local domain discrepancy; 3) a global generative discrepancy\nmetric (GGDM) is presented, such that both the global and local discrepancy can\nbe effectively and positively reduced; 4) a simplified version of MCTL called\nMCTL-S is presented under a perfect domain generation assumption for more\ngeneric learning scenario. Experiments on a number of benchmark visual transfer\ntasks demonstrate the superiority of the proposed manifold criterion guided\ngenerative transfer method, by comparing with other state-of-the-art methods.\nThe source code is available in https://github.com/wangshanshanCQU/MCTL.","url_abs":"http://arxiv.org/abs/1903.10211v1","url_pdf":"http://arxiv.org/pdf/1903.10211v1.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":"manifold-criterion-guided-transfer-learning","repo_url":"https://github.com/wangshanshanCQU/MCTL","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":null}],"tasks":[{"task_slug":"domain-adaptation","task_name":"Domain Adaptation"},{"task_slug":"transfer-learning","task_name":"Transfer Learning"},{"task_slug":"unsupervised-domain-adaptation","task_name":"Unsupervised Domain Adaptation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}