{"id":172,"date":"2006-03-05T12:50:38","date_gmt":"2006-03-05T18:50:38","guid":{"rendered":"http:\/\/hunch.net\/?p=172"},"modified":"2006-03-05T12:59:03","modified_gmt":"2006-03-05T18:59:03","slug":"structural-learning","status":"publish","type":"post","link":"https:\/\/hunch.net\/?p=172","title":{"rendered":"&#8220;Structural&#8221; Learning"},"content":{"rendered":"<p><a href=\"http:\/\/www.cis.upenn.edu\/~pereira\/\">Fernando Pereira<\/a> <a href=\"https:\/\/hunch.net\/?p=153#comments\">pointed out<\/a> Ando and <a href=\"http:\/\/www-cs-students.stanford.edu\/~tzhang\/\">Zhang<\/a>&#8216;s <a href=\"http:\/\/www-cs-students.stanford.edu\/~tzhang\/papers\/jmlr05_semisup.pdf\">paper<\/a> on &#8220;structural&#8221; learning.  Structural learning is multitask learning on subproblems created from unlabeled data.  <\/p>\n<p>The basic idea is to take a look at the unlabeled data and create many supervised problems.  On text data, which they test on, these subproblems might be of the form &#8220;Given surrounding words predict the middle word&#8221;.   The hope here is that successfully predicting on these subproblems is relevant to the prediction of your core problem.<\/p>\n<p>In the long run, the precise mechanism used (essentially, linear predictors with parameters tied by a common matrix) and the precise problems formed may not be critical.  What seems critical is that the hope is realized: the technique provides a significant edge in practice.<\/p>\n<p>Some basic questions about this approach are:<\/p>\n<ol>\n<li>Are there effective automated mechanisms for creating the subproblems?<\/li>\n<li>Is it necessary to use a shared representation?<\/li>\n<\/ol>\n","protected":false},"excerpt":{"rendered":"<p>Fernando Pereira pointed out Ando and Zhang&#8216;s paper on &#8220;structural&#8221; learning. Structural learning is multitask learning on subproblems created from unlabeled data. The basic idea is to take a look at the unlabeled data and create many supervised problems. On text data, which they test on, these subproblems might be of the form &#8220;Given surrounding &hellip; <\/p>\n<p class=\"link-more\"><a href=\"https:\/\/hunch.net\/?p=172\" class=\"more-link\">Continue reading<span class=\"screen-reader-text\"> &#8220;&#8220;Structural&#8221; Learning&#8221;<\/span><\/a><\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[29,18,9],"tags":[],"class_list":["post-172","post","type-post","status-publish","format-standard","hentry","category-machine-learning","category-papers","category-semisupervised"],"_links":{"self":[{"href":"https:\/\/hunch.net\/index.php?rest_route=\/wp\/v2\/posts\/172","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/hunch.net\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/hunch.net\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/hunch.net\/index.php?rest_route=\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/hunch.net\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=172"}],"version-history":[{"count":0,"href":"https:\/\/hunch.net\/index.php?rest_route=\/wp\/v2\/posts\/172\/revisions"}],"wp:attachment":[{"href":"https:\/\/hunch.net\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=172"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/hunch.net\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=172"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/hunch.net\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=172"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}