We examine a class of stochastic deep learning models with a tractable method to compute information-theoretic quantities. Our contributions are three-fold: (i) We show how entropies and mutual informations can be derived from heuristic statistical physics methods, under the assumption that weight matrices are independent and orthogonally-invariant. (ii) We extend particular cases in which this result is known to be rigorously exact by providing a proof for two-layers networks with Gaussian random weights, using the recently introduced adaptive interpolation method. (iii) We propose an experiment framework with generative models of synthetic datasets, on which we train deep neural networks with a weight constraint designed so that the assumption in (i) is verified during learning. We study the behavior of entropies and mutual information throughout learning and conclude that, in the proposed setting, the relationship between compression and generalization remains elusive.<\/p>\n","protected":false},"excerpt":{"rendered":"
We examine a class of stochastic deep learning models with a tractable method to compute information-theoretic quantities. Our contributions are three-fold: (i) We show how entropies and mutual informations can be derived from heuristic statistical physics methods, under the assumption that weight matrices are independent and orthogonally-invariant. (ii) We extend particular cases in which this […]<\/p>\n","protected":false},"featured_media":0,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","msr-author-ordering":[{"type":"text","value":"Marylou Gabri\u00e9","user_id":0},{"type":"user_nicename","value":"Andre Manoel","user_id":"40504"},{"type":"text","value":"Cl\u00e9ment Luneau","user_id":0},{"type":"text","value":"Jean Barbier","user_id":0},{"type":"text","value":"Nicolas Macris","user_id":0},{"type":"text","value":"Florent Krzakala","user_id":0},{"type":"text","value":"Lenka Zdeborov\u00e1","user_id":0}],"msr_publishername":"","msr_publisher_other":"","msr_booktitle":"","msr_chapter":"","msr_edition":"","msr_editors":"","msr_how_published":"","msr_isbn":"","msr_issue":"","msr_journal":"","msr_number":"","msr_organization":"","msr_pages_string":"1821\u20131831","msr_page_range_start":"1821","msr_page_range_end":"1831","msr_series":"","msr_volume":"","msr_copyright":"","msr_conference_name":"2018 Neural Information Processing Systems","msr_doi":"","msr_arxiv_id":"","msr_mag_id":"2963144505","msr_other_authors":"","msr_other_contributors":"","msr_speaker":"","msr_award":"","msr_affiliation":"","msr_institution":"","msr_host":"","msr_version":"","msr_duration":"","msr_release_tracker_id":"","msr_highlight_type":"","msr_date_display_format":"","msr_main_download_label":"","msr_external_link_label":"","msr_doi_label":"","msr_published_date":"2018-12-31","msr_startdate":"","msr_presentation_date":"","msr_highlight_text":"","msr_notes":"","msr_longbiography":"","msr_publicationurl":"","msr_external_url":"","msr_secondary_video_url":"","msr_conference_url":"","msr_journal_url":"","msr_year":2018,"msr_month":12,"msr_day":31,"msr_microsoftintellectualproperty":false,"msr_pub_id":"","msr_publication_uploader":[{"type":"url","viewUrl":"false","id":false,"title":"https:\/\/papers.nips.cc\/paper\/2018\/file\/6d0f846348a856321729a2f36734d1a7-Paper.pdf","label_id":243132,"label":0},{"type":"url","viewUrl":"false","id":false,"title":"https:\/\/proceedings.neurips.cc\/paper\/2018\/file\/6d0f846348a856321729a2f36734d1a7-Paper.pdf","label_id":243132,"label":0},{"type":"url","viewUrl":"false","id":false,"title":"http:\/\/papers.nips.cc\/paper\/7453-entropy-and-mutual-information-in-models-of-deep-neural-networks.pdf","label_id":243109,"label":0},{"type":"url","viewUrl":"false","id":false,"title":"https:\/\/dblp.uni-trier.de\/db\/conf\/nips\/nips2018.html#GabrieMLBMKZ18","label_id":243109,"label":0},{"type":"url","viewUrl":"false","id":false,"title":"https:\/\/proceedings.neurips.cc\/paper\/7453-entropy-and-mutual-information-in-models-of-deep-neural-networks","label_id":243109,"label":0}],"msr_related_uploader":[],"msr_original_fields_of_study":["Deep learning","Computer science","Artificial intelligence","Generative grammar","Algorithm","Deep neural networks","Mutual information","Gaussian","Entropy (information theory)","Entropy (statistical thermodynamics)","Entropy (energy dispersal)","Entropy (classical thermodynamics)","Entropy (arrow of time)","Entropy (order and disorder)"],"msr_s2_paper_id":"","msr_s2_pdf_url":"","msr_citation_count_updated":"","msr_citation_count":0,"msr_influential_citations":0,"msr_reference_count":0,"msr_s2_open_access":false,"msr_s2_author_ids":[],"msr_pub_ids":[],"msr_hide_image_in_river":0,"footnotes":""},"msr-research-highlight":[],"research-area":[13556],"msr-publication-type":[193716],"msr-publisher":[],"msr-publication-cta":[],"msr-focus-area":[],"msr-locale":[268875],"msr-post-option":[],"msr-field-of-study":[246904,246694,246691,246658,248335,258151,258148,258145,257761,258154,258142,250777,246748,249943],"msr-conference":[],"msr-journal":[],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-760711","msr-research-item","type-msr-research-item","status-publish","hentry","msr-research-area-artificial-intelligence","msr-locale-en_us","msr-field-of-study-algorithm","msr-field-of-study-artificial-intelligence","msr-field-of-study-computer-science","msr-field-of-study-deep-learning","msr-field-of-study-deep-neural-networks","msr-field-of-study-entropy-arrow-of-time","msr-field-of-study-entropy-classical-thermodynamics","msr-field-of-study-entropy-energy-dispersal","msr-field-of-study-entropy-information-theory","msr-field-of-study-entropy-order-and-disorder","msr-field-of-study-entropy-statistical-thermodynamics","msr-field-of-study-gaussian","msr-field-of-study-generative-grammar","msr-field-of-study-mutual-information"],"msr_publishername":"","msr_edition":"","msr_affiliation":"","msr_published_date":"2018-12-31","msr_host":"","msr_duration":"","msr_version":"","msr_speaker":"","msr_other_contributors":"","msr_booktitle":"","msr_pages_string":"1821\u20131831","msr_chapter":"","msr_isbn":"","msr_journal":"","msr_volume":"","msr_number":"","msr_editors":"","msr_series":"","msr_issue":"","msr_organization":"","msr_how_published":"","msr_notes":"","msr_highlight_text":"","msr_release_tracker_id":"","msr_original_fields_of_study":"","msr_download_urls":"","msr_external_url":"","msr_secondary_video_url":"","msr_longbiography":"","msr_microsoftintellectualproperty":0,"msr_main_download":"","msr_publicationurl":"","msr_doi":"","msr_publication_uploader":[{"type":"url","viewUrl":"false","id":"false","title":"https:\/\/papers.nips.cc\/paper\/2018\/file\/6d0f846348a856321729a2f36734d1a7-Paper.pdf","label_id":"243132","label":0},{"type":"url","viewUrl":"false","id":"false","title":"https:\/\/proceedings.neurips.cc\/paper\/2018\/file\/6d0f846348a856321729a2f36734d1a7-Paper.pdf","label_id":"243132","label":0},{"type":"url","viewUrl":"false","id":"false","title":"http:\/\/papers.nips.cc\/paper\/7453-entropy-and-mutual-information-in-models-of-deep-neural-networks.pdf","label_id":"243109","label":0},{"type":"url","viewUrl":"false","id":"false","title":"https:\/\/dblp.uni-trier.de\/db\/conf\/nips\/nips2018.html#GabrieMLBMKZ18","label_id":"243109","label":0},{"type":"url","viewUrl":"false","id":"false","title":"https:\/\/proceedings.neurips.cc\/paper\/7453-entropy-and-mutual-information-in-models-of-deep-neural-networks","label_id":"243109","label":0}],"msr_related_uploader":[],"msr_citation_count":0,"msr_citation_count_updated":"","msr_s2_paper_id":"","msr_influential_citations":0,"msr_reference_count":0,"msr_arxiv_id":"","msr_s2_author_ids":[],"msr_s2_open_access":false,"msr_s2_pdf_url":null,"msr_attachments":[],"msr-author-ordering":[{"type":"text","value":"Marylou Gabri\u00e9","user_id":0,"rest_url":false},{"type":"user_nicename","value":"Andre Manoel","user_id":40504,"rest_url":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Andre Manoel"},{"type":"text","value":"Cl\u00e9ment Luneau","user_id":0,"rest_url":false},{"type":"text","value":"Jean Barbier","user_id":0,"rest_url":false},{"type":"text","value":"Nicolas Macris","user_id":0,"rest_url":false},{"type":"text","value":"Florent Krzakala","user_id":0,"rest_url":false},{"type":"text","value":"Lenka Zdeborov\u00e1","user_id":0,"rest_url":false}],"msr_impact_theme":[],"msr_research_lab":[],"msr_event":[],"msr_group":[],"msr_project":[],"publication":[],"video":[],"msr-tool":[],"msr_publication_type":"inproceedings","related_content":[],"_links":{"self":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/760711","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item"}],"about":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-research-item"}],"version-history":[{"count":1,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/760711\/revisions"}],"predecessor-version":[{"id":760714,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/760711\/revisions\/760714"}],"wp:attachment":[{"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/media?parent=760711"}],"wp:term":[{"taxonomy":"msr-research-highlight","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-research-highlight?post=760711"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=760711"},{"taxonomy":"msr-publication-type","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-publication-type?post=760711"},{"taxonomy":"msr-publisher","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-publisher?post=760711"},{"taxonomy":"msr-publication-cta","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-publication-cta?post=760711"},{"taxonomy":"msr-focus-area","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-focus-area?post=760711"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=760711"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=760711"},{"taxonomy":"msr-field-of-study","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-field-of-study?post=760711"},{"taxonomy":"msr-conference","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-conference?post=760711"},{"taxonomy":"msr-journal","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-journal?post=760711"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=760711"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/cm-edgetun.pages.dev\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=760711"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}