{"id":558135,"date":"2018-12-26T00:17:35","date_gmt":"2018-12-26T08:17:35","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/?post_type=msr-project&p=558135"},"modified":"2021-07-04T20:15:11","modified_gmt":"2021-07-05T03:15:11","slug":"dual-learning","status":"publish","type":"msr-project","link":"https:\/\/www.microsoft.com\/en-us\/research\/project\/dual-learning\/","title":{"rendered":"Dual Learning"},"content":{"rendered":"

Introduction to Dual Learning<\/strong>
\nMany AI tasks are emerged in dual forms, e.g., English-to-French translation vs. French-to-English translation, speech recognition vs. text to speech, question answering vs. question generation, and image classification vs. image generation. While structural duality is common in AI, most learning algorithms have not exploited it in learning\/inference. We propose a new learning paradigm, dual learning,\u00a0 which leverages the primal-dual structure of AI tasks to obtain effective feedback or regularization signals to enhance the learning\/inference process. Dual learning has been studied in different learning settings, including unsupervised\/supervised\/semi-supervised\/transfer settings, and applied to different applications, including machine translation, sentimental analysis, image classification\/generation, question answering\/generation …<\/p>\n

Tutorial\u00a0and code<\/strong>
\nThe book on Dual Learning (opens in new tab)<\/span><\/a> is published by Springer!<\/p>\n

\"\" (opens in new tab)<\/span><\/a><\/p>\n

Tutorial\u00a0and code<\/strong>
\nTutorial (opens in new tab)<\/span><\/a> at IJCAI 2019
\nTutorial on dual learning (opens in new tab)<\/span><\/a> at ACML 2018<\/span>
\nDual Supervised Learning for image classification\/generation and sentiment analysis, [<\/span>Code@Github (opens in new tab)<\/span><\/a>]<\/span><\/p>\n

Our papers<\/strong>
\nYiren Wang, Yingce Xia, Tianyu He, Fei Tian, Tao Qin, ChengXiang Zhai, Tie-Yan Liu, <\/span>Multi-Agent Dual Learning (opens in new tab)<\/span><\/a>, <\/span>ICLR<\/strong> 2019.<\/span>
\nYingce Xia, Xu Tan, Fei Tian, Tao Qin, Nenghai Yu, and Tie-Yan Liu, <\/span>Model-Level Dual Learning (opens in new tab)<\/span><\/a>, <\/span>ICML<\/strong> 2018.<\/span>
\nHany Hassan, Anthony Aue, Chang Chen, Vishal Chowdhary, Jonathan Clark, Christian Federmann, Xuedong Huang, Marcin Junczys-Dowmunt, William Lewis, Mu Li, Shujie Liu, Tie-Yan Liu, Renqian Luo, Arul Menezes, Tao Qin, Frank Seide, Xu Tan, Fei Tian, Lijun Wu, Shuangzhi Wu, Yingce Xia, Dongdong Zhang, Zhirui Zhang, Ming Zhou, <\/span>Achieving Human Parity on Automatic Chinese to English News Translation (opens in new tab)<\/span><\/a>, arXiv 2018.<\/span>
\nJianxin Lin, Yingce Xia, Tao Qin, Zhibo Chen, and Tie-Yan Liu, <\/span>Conditional Image-to-Image Translation (opens in new tab)<\/span><\/a>, <\/span>CVPR<\/strong> 2018.<\/span>
\nYijun Wang, Yingce Xia, Li Zhao, Jiang Bian, Tao Qin, Guiquan Liu, and Tie-Yan Liu, <\/span>Dual Transfer Learning for Neural Machine Translation with Marginal Distribution Regularization (opens in new tab)<\/span><\/a>, <\/span>AAAI<\/strong> 2018.<\/span>
\nDuyu Tang, Nan Duan, Tao Qin, Zhao Yan, and Ming Zhou, Question Answering and Question Generation as Dual Tasks, arXiv 2017.<\/span>
\nYingce Xia, Tao Qin, Wei Chen, Jiang Bian, Nenghai Yu and Tie-Yan Liu, <\/span>Dual Supervised Learning (opens in new tab)<\/span><\/a>,\u00a0 <\/span>ICML <\/b>2017<\/span>.<\/b>
\nYingce Xia, Jiang Bian, Tao Qin, Tie-Yan Liu, <\/span>Dual Inference for Machine Learning (opens in new tab)<\/span><\/a>, <\/span>IJCAI<\/strong> 2017.
\nDi He, Yingce Xia, Tao Qin, Liwei Wang, Nenghai Yu, Tie-Yan Liu, and Wei-Ying Ma, Dual Learning for Machine Translation (opens in new tab)<\/span><\/a>, NIPS 2016.<\/span>\"\"<\/span><\/p>\n

 <\/p>\n

 <\/p>\n

 <\/p>\n

More papers<\/strong> (opens in new tab)<\/span><\/a><\/p>\n

 <\/p>\n","protected":false},"excerpt":{"rendered":"

Introduction to Dual Learning Many AI tasks are emerged in dual forms, e.g., English-to-French translation vs. French-to-English translation, speech recognition vs. text to speech, question answering vs. question generation, and image classification vs. image generation. While structural duality is common in AI, most learning algorithms have not exploited it in learning\/inference. We propose a new […]<\/p>\n","protected":false},"featured_media":0,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"footnotes":""},"research-area":[13556],"msr-locale":[268875],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-558135","msr-project","type-msr-project","status-publish","hentry","msr-research-area-artificial-intelligence","msr-locale-en_us","msr-archive-status-active"],"msr_project_start":"","related-publications":[],"related-downloads":[],"related-videos":[],"related-groups":[],"related-events":[],"related-opportunities":[],"related-posts":[],"related-articles":[],"tab-content":[],"slides":[],"related-researchers":[{"type":"user_nicename","display_name":"Yingce Xia","user_id":37784,"people_section":"Section name 1","alias":"yinxia"},{"type":"user_nicename","display_name":"Tao Qin","user_id":33871,"people_section":"Section name 1","alias":"taoqin"},{"type":"user_nicename","display_name":"Li Zhao","user_id":36152,"people_section":"Section name 1","alias":"lizo"},{"type":"user_nicename","display_name":"Tie-Yan Liu","user_id":34431,"people_section":"Section name 1","alias":"tyliu"}],"msr_research_lab":[199560],"msr_impact_theme":[],"_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/558135"}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-project"}],"version-history":[{"count":9,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/558135\/revisions"}],"predecessor-version":[{"id":757894,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-project\/558135\/revisions\/757894"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=558135"}],"wp:term":[{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=558135"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=558135"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=558135"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=558135"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}