{"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/peephole-predicting-network-performance","title":"Peephole: Predicting Network Performance Before Training","arxiv_id":"1712.03351","date":"2017-12-09","proceeding":null,"authors":["Boyang Deng","Junjie Yan","Dahua Lin"],"abstract":"The quest for performant networks has been a significant force that drives\nthe advancements of deep learning in recent years. While rewarding, improving\nnetwork design has never been an easy journey. The large design space combined\nwith the tremendous cost required for network training poses a major obstacle\nto this endeavor. In this work, we propose a new approach to this problem,\nnamely, predicting the performance of a network before training, based on its\narchitecture. Specifically, we develop a unified way to encode individual\nlayers into vectors and bring them together to form an integrated description\nvia LSTM. Taking advantage of the recurrent network's strong expressive power,\nthis method can reliably predict the performances of various network\narchitectures. Our empirical studies showed that it not only achieved accurate\npredictions but also produced consistent rankings across datasets -- a key\ndesideratum in performance prediction.","url_abs":"http://arxiv.org/abs/1712.03351v1","url_pdf":"http://arxiv.org/pdf/1712.03351v1.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":"peephole-predicting-network-performance","repo_url":"https://github.com/pdefraene/cgp_cnn_predictors","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}}],"tasks":[],"methods":[{"method_slug":"lstm","method_name":"LSTM"},{"method_slug":"sigmoid-activation","method_name":"Sigmoid Activation"},{"method_slug":"tanh-activation","method_name":"Tanh Activation"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1712.03351","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}