{"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/flattened-convolutional-neural-networks-for","title":"Flattened Convolutional Neural Networks for Feedforward Acceleration","arxiv_id":"1412.5474","date":"2014-12-17","proceeding":null,"authors":["Jonghoon Jin","Aysegul Dundar","Eugenio Culurciello"],"abstract":"We present flattened convolutional neural networks that are designed for fast\nfeedforward execution. The redundancy of the parameters, especially weights of\nthe convolutional filters in convolutional neural networks has been extensively\nstudied and different heuristics have been proposed to construct a low rank\nbasis of the filters after training. In this work, we train flattened networks\nthat consist of consecutive sequence of one-dimensional filters across all\ndirections in 3D space to obtain comparable performance as conventional\nconvolutional networks. We tested flattened model on different datasets and\nfound that the flattened layer can effectively substitute for the 3D filters\nwithout loss of accuracy. The flattened convolution pipelines provide around\ntwo times speed-up during feedforward pass compared to the baseline model due\nto the significant reduction of learning parameters. Furthermore, the proposed\nmethod does not require efforts in manual tuning or post processing once the\nmodel is trained.","url_abs":"http://arxiv.org/abs/1412.5474v4","url_pdf":"http://arxiv.org/pdf/1412.5474v4.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":"flattened-convolutional-neural-networks-for","repo_url":"https://github.com/jhjin/flattened-cnn","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"torch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"flattened-convolutional-neural-networks-for","repo_url":"https://github.com/Liangrj5/Flatten_Net","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[],"methods":[{"method_slug":"convolution","method_name":"Convolution"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1412.5474","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}