{"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/an-analysis-of-unsupervised-pre-training-in","title":"An Analysis of Unsupervised Pre-training in Light of Recent Advances","arxiv_id":"1412.6597","date":"2014-12-20","proceeding":null,"authors":["Tom Le Paine","Pooya Khorrami","Wei Han","Thomas S. Huang"],"abstract":"Convolutional neural networks perform well on object recognition because of a\nnumber of recent advances: rectified linear units (ReLUs), data augmentation,\ndropout, and large labelled datasets. Unsupervised data has been proposed as\nanother way to improve performance. Unfortunately, unsupervised pre-training is\nnot used by state-of-the-art methods leading to the following question: Is\nunsupervised pre-training still useful given recent advances? If so, when? We\nanswer this in three parts: we 1) develop an unsupervised method that\nincorporates ReLUs and recent unsupervised regularization techniques, 2)\nanalyze the benefits of unsupervised pre-training compared to data augmentation\nand dropout on CIFAR-10 while varying the ratio of unsupervised to supervised\nsamples, 3) verify our findings on STL-10. We discover unsupervised\npre-training, as expected, helps when the ratio of unsupervised to supervised\nsamples is high, and surprisingly, hurts when the ratio is low. We also use\nunsupervised pre-training with additional color augmentation to achieve near\nstate-of-the-art performance on STL-10.","url_abs":"http://arxiv.org/abs/1412.6597v4","url_pdf":"http://arxiv.org/pdf/1412.6597v4.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":"an-analysis-of-unsupervised-pre-training-in","repo_url":"https://github.com/ifp-uiuc/an-analysis-of-unsupervised-pre-training-iclr-2015","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":null},{"paper_slug":"an-analysis-of-unsupervised-pre-training-in","repo_url":"https://github.com/ifp-uiuc/anna","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":"data-augmentation","task_name":"Data Augmentation"},{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"object-recognition","task_name":"Object Recognition"},{"task_slug":"unsupervised-pre-training","task_name":"Unsupervised Pre-training"}],"methods":[{"method_slug":"dropout","method_name":"Dropout"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/image-classification-on-cifar-10","task":"Image Classification","dataset":"CIFAR-10","model":"An Analysis of Unsupervised Pre-training in Light of Recent Advances","rank_in_archive_order":222,"of":265,"metrics":{"Percentage correct":"86.7"},"uses_additional_data":false},{"leaderboard":"/sota/image-classification-on-stl-10","task":"Image Classification","dataset":"STL-10","model":"An Analysis of Unsupervised Pre-training in Light of Recent Advances","rank_in_archive_order":91,"of":117,"metrics":{"Percentage correct":"70.2"},"uses_additional_data":false}],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}