{"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/the-deep-weight-prior","title":"The Deep Weight Prior","arxiv_id":"1810.06943","date":"2018-10-16","proceeding":"ICLR 2019 5","authors":["Andrei Atanov","Arsenii Ashukha","Kirill Struminsky","Dmitry Vetrov","Max Welling"],"abstract":"Bayesian inference is known to provide a general framework for incorporating\nprior knowledge or specific properties into machine learning models via\ncarefully choosing a prior distribution. In this work, we propose a new type of\nprior distributions for convolutional neural networks, deep weight prior (DWP),\nthat exploit generative models to encourage a specific structure of trained\nconvolutional filters e.g., spatial correlations of weights. We define DWP in\nthe form of an implicit distribution and propose a method for variational\ninference with such type of implicit priors. In experiments, we show that DWP\nimproves the performance of Bayesian neural networks when training data are\nlimited, and initialization of weights with samples from DWP accelerates\ntraining of conventional convolutional neural networks.","url_abs":"http://arxiv.org/abs/1810.06943v6","url_pdf":"http://arxiv.org/pdf/1810.06943v6.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":"the-deep-weight-prior","repo_url":"https://github.com/matyushinleonid/WeightPrior","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"bayesian-inference","task_name":"Bayesian Inference"},{"task_slug":"variational-inference","task_name":"Variational Inference"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1810.06943","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}