{"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/compressing-neural-networks-with-the-hashing","title":"Compressing Neural Networks with the Hashing Trick","arxiv_id":"1504.04788","date":"2015-04-19","proceeding":null,"authors":["Wenlin Chen","James T. Wilson","Stephen Tyree","Kilian Q. Weinberger","Yixin Chen"],"abstract":"As deep nets are increasingly used in applications suited for mobile devices,\na fundamental dilemma becomes apparent: the trend in deep learning is to grow\nmodels to absorb ever-increasing data set sizes; however mobile devices are\ndesigned with very little memory and cannot store such large models. We present\na novel network architecture, HashedNets, that exploits inherent redundancy in\nneural networks to achieve drastic reductions in model sizes. HashedNets uses a\nlow-cost hash function to randomly group connection weights into hash buckets,\nand all connections within the same hash bucket share a single parameter value.\nThese parameters are tuned to adjust to the HashedNets weight sharing\narchitecture with standard backprop during training. Our hashing procedure\nintroduces no additional memory overhead, and we demonstrate on several\nbenchmark data sets that HashedNets shrink the storage requirements of neural\nnetworks substantially while mostly preserving generalization performance.","url_abs":"http://arxiv.org/abs/1504.04788v1","url_pdf":"http://arxiv.org/pdf/1504.04788v1.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":"compressing-neural-networks-with-the-hashing","repo_url":"https://gitlab.com/anon-dynamic-reparam/iclr2019-dynamic-reparam","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1504.04788","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}