{"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/aet-vs-aed-unsupervised-representation","title":"AET vs. AED: Unsupervised Representation Learning by Auto-Encoding Transformations rather than Data","arxiv_id":"1901.04596","date":"2019-01-14","proceeding":"CVPR 2019 6","authors":["Liheng Zhang","Guo-Jun Qi","Liqiang Wang","Jiebo Luo"],"abstract":"The success of deep neural networks often relies on a large amount of labeled\nexamples, which can be difficult to obtain in many real scenarios. To address\nthis challenge, unsupervised methods are strongly preferred for training neural\nnetworks without using any labeled data. In this paper, we present a novel\nparadigm of unsupervised representation learning by Auto-Encoding\nTransformation (AET) in contrast to the conventional Auto-Encoding Data (AED)\napproach. Given a randomly sampled transformation, AET seeks to predict it\nmerely from the encoded features as accurately as possible at the output end.\nThe idea is the following: as long as the unsupervised features successfully\nencode the essential information about the visual structures of original and\ntransformed images, the transformation can be well predicted. We will show that\nthis AET paradigm allows us to instantiate a large variety of transformations,\nfrom parameterized, to non-parameterized and GAN-induced ones. Our experiments\nshow that AET greatly improves over existing unsupervised approaches, setting\nnew state-of-the-art performances being greatly closer to the upper bounds by\ntheir fully supervised counterparts on CIFAR-10, ImageNet and Places datasets.","url_abs":"http://arxiv.org/abs/1901.04596v2","url_pdf":"http://arxiv.org/pdf/1901.04596v2.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":"aet-vs-aed-unsupervised-representation","repo_url":"https://github.com/maple-research-lab/AET","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"representation-learning","task_name":"Representation Learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1901.04596","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}