{"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/when-unsupervised-domain-adaptation-meets","title":"When Unsupervised Domain Adaptation Meets Tensor Representations","arxiv_id":"1707.05956","date":"2017-07-19","proceeding":"ICCV 2017 10","authors":["Hao Lu","Lei Zhang","Zhiguo Cao","Wei Wei","Ke Xian","Chunhua Shen","Anton Van Den Hengel"],"abstract":"Domain adaption (DA) allows machine learning methods trained on data sampled\nfrom one distribution to be applied to data sampled from another. It is thus of\ngreat practical importance to the application of such methods. Despite the fact\nthat tensor representations are widely used in Computer Vision to capture\nmulti-linear relationships that affect the data, most existing DA methods are\napplicable to vectors only. This renders them incapable of reflecting and\npreserving important structure in many problems. We thus propose here a\nlearning-based method to adapt the source and target tensor representations\ndirectly, without vectorization. In particular, a set of alignment matrices is\nintroduced to align the tensor representations from both domains into the\ninvariant tensor subspace. These alignment matrices and the tensor subspace are\nmodeled as a joint optimization problem and can be learned adaptively from the\ndata using the proposed alternative minimization scheme. Extensive experiments\nshow that our approach is capable of preserving the discriminative power of the\nsource domain, of resisting the effects of label noise, and works effectively\nfor small sample sizes, and even one-shot DA. We show that our method\noutperforms the state-of-the-art on the task of cross-domain visual recognition\nin both efficacy and efficiency, and particularly that it outperforms all\ncomparators when applied to DA of the convolutional activations of deep\nconvolutional networks.","url_abs":"http://arxiv.org/abs/1707.05956v1","url_pdf":"http://arxiv.org/pdf/1707.05956v1.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":"when-unsupervised-domain-adaptation-meets","repo_url":"https://github.com/poppinace/TAISL","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":"unsupervised-domain-adaptation","task_name":"Unsupervised Domain Adaptation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1707.05956","atlas_url":"https://app.syntology.ai/?focus=1707.05956","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}