{"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/avoiding-pathologies-in-very-deep-networks","title":"Avoiding pathologies in very deep networks","arxiv_id":"1402.5836","date":"2014-02-24","proceeding":null,"authors":["David Duvenaud","Oren Rippel","Ryan P. Adams","Zoubin Ghahramani"],"abstract":"Choosing appropriate architectures and regularization strategies for deep\nnetworks is crucial to good predictive performance. To shed light on this\nproblem, we analyze the analogous problem of constructing useful priors on\ncompositions of functions. Specifically, we study the deep Gaussian process, a\ntype of infinitely-wide, deep neural network. We show that in standard\narchitectures, the representational capacity of the network tends to capture\nfewer degrees of freedom as the number of layers increases, retaining only a\nsingle degree of freedom in the limit. We propose an alternate network\narchitecture which does not suffer from this pathology. We also examine deep\ncovariance functions, obtained by composing infinitely many feature transforms.\nLastly, we characterize the class of models obtained by performing dropout on\nGaussian processes.","url_abs":"http://arxiv.org/abs/1402.5836v3","url_pdf":"http://arxiv.org/pdf/1402.5836v3.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":"avoiding-pathologies-in-very-deep-networks","repo_url":"https://github.com/duvenaud/deep-limits","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"avoiding-pathologies-in-very-deep-networks","repo_url":"https://github.com/duvenaud/additive-gps","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"GPL-2.0"}}],"tasks":[{"task_slug":"gaussian-processes","task_name":"Gaussian Processes"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1402.5836","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}