{"id":118539,"date":"2021-11-01T03:20:18","date_gmt":"2021-10-31T18:20:18","guid":{"rendered":"http:\/\/175.125.95.178\/ai-in-communication\/18539\/"},"modified":"2026-05-03T04:35:49","modified_gmt":"2026-05-02T19:35:49","slug":"18539","status":"publish","type":"ai-in-communication","link":"http:\/\/ee.presscat.kr\/en\/ai-in-communication\/18539\/","title":{"rendered":"Unsupervised Embedding Adaptation via Early-Stage Feature Reconstruction for Few-shot Classification, ICML 2021"},"content":{"rendered":"<p align=\"left\" style=\"text-align:left\"><span style=\"font-size:10pt\"><span style=\"background:white\"><span style=\"line-height:normal\"><span><span><span><span lang=\"EN-US\" style=\",sans-serif\"><span style=\"color:black\">We propose unsupervised embedding adaptation for the downstream few-shot classification task.<\/span><\/span><\/span><\/span><\/span><\/span><\/span><\/span><\/p>\n<p align=\"left\" style=\"text-align:left\"><span style=\"font-size:10pt\"><span style=\"background:white\"><span style=\"line-height:normal\"><span><span><span><span lang=\"EN-US\" style=\",sans-serif\"><span style=\"color:black\">Based on findings that deep neural networks learn to generalize before memorizing, we develop Early-Stage Feature Reconstruction (ESFR) &#8212; a novel adaptation scheme with feature reconstruction and dimensionality-driven early stopping that finds generalizable features.<\/span><\/span><\/span><\/span><\/span><\/span><\/span><\/span><\/p>\n<p align=\"left\" style=\"text-align:left\"><span style=\"font-size:10pt\"><span style=\"background:white\"><span style=\"line-height:normal\"><span><span><span><span lang=\"EN-US\" style=\",sans-serif\"><span style=\"color:black\">Incorporating ESFR consistently improves the performance of baseline methods on all standard settings, including the recently proposed transductive method.<\/span><\/span><\/span><\/span><\/span><\/span><\/span><\/span><\/p>\n<p><span lang=\"EN-US\" style=\"font-size:10.0pt\"><span style=\"line-height:107%\"><span style=\",sans-serif\"><span style=\"color:black\">ESFR used in conjunction with the transductive method further achieves state-of-the-art performance on mini-ImageNet, tiered-ImageNet, and CUB; especially with 1.2%~2.0% improvements in accuracy over the previous best performing method on 1-shot setting. <\/span><\/span><\/span><\/span><\/p>\n<p><div class=\"\"><img decoding=\"async\" class=\"\" src=\"\/wp-content\/uploads\/drupal\/\uc815\uc138\uc601\uad50\uc218\ub2d81.png\" alt=\"\" title=\"\"><\/div><\/p>\n","protected":false},"excerpt":{"rendered":"<p>674<\/p>\n","protected":false},"featured_media":0,"template":"","class_list":["post-118539","ai-in-communication","type-ai-in-communication","status-publish","hentry"],"acf":[],"_links":{"self":[{"href":"http:\/\/ee.presscat.kr\/en\/wp-json\/wp\/v2\/ai-in-communication\/118539","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/ee.presscat.kr\/en\/wp-json\/wp\/v2\/ai-in-communication"}],"about":[{"href":"http:\/\/ee.presscat.kr\/en\/wp-json\/wp\/v2\/types\/ai-in-communication"}],"wp:attachment":[{"href":"http:\/\/ee.presscat.kr\/en\/wp-json\/wp\/v2\/media?parent=118539"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}