{"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/learning-deep-representations-using","title":"Learning Deep Representations Using Convolutional Auto-encoders with Symmetric Skip Connections","arxiv_id":"1611.09119","date":"2016-11-28","proceeding":null,"authors":["Jianfeng Dong","Xiao-Jiao Mao","Chunhua Shen","Yu-Bin Yang"],"abstract":"Unsupervised pre-training was a critical technique for training deep neural\nnetworks years ago. With sufficient labeled data and modern training\ntechniques, it is possible to train very deep neural networks from scratch in a\npurely supervised manner nowadays. However, unlabeled data is easier to obtain\nand usually of very large scale. How to make use of them better to help\nsupervised learning is still a well-valued topic. In this paper, we investigate\nconvolutional denoising auto-encoders to show that unsupervised pre-training\ncan still improve the performance of high-level image related tasks such as\nimage classification and semantic segmentation. The architecture we use is a\nconvolutional auto-encoder network with symmetric shortcut connections. We\nempirically show that symmetric shortcut connections are very important for\nlearning abstract representations via image reconstruction. When no extra\nunlabeled data are available, unsupervised pre-training with our network can\nregularize the supervised training and therefore lead to better generalization\nperformance. With the help of unsupervised pre-training, our method achieves\nvery competitive results in image classification using very simple\nall-convolution networks. When labeled data are limited but extra unlabeled\ndata are available, our method achieves good results in several semi-supervised\nlearning tasks.","url_abs":"http://arxiv.org/abs/1611.09119v2","url_pdf":"http://arxiv.org/pdf/1611.09119v2.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":"learning-deep-representations-using","repo_url":"https://github.com/anushkayadav/Denoising_cifar10","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"denoising","task_name":"Denoising"},{"task_slug":"classification","task_name":"General Classification"},{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"image-reconstruction","task_name":"Image Reconstruction"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"},{"task_slug":"unsupervised-pre-training","task_name":"Unsupervised Pre-training"},{"task_slug":"image-classification","task_name":"image-classification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}