{"entities":{"Q1983625":{"pageid":1994367,"ns":120,"title":"Item:Q1983625","lastrevid":72206538,"modified":"2026-04-14T03:18:34Z","type":"item","id":"Q1983625","labels":{"en":{"language":"en","value":"Over-parametrized deep neural networks minimizing the empirical risk do not generalize well"}},"descriptions":{"en":{"language":"en","value":"scientific article; zbMATH DE number 7394101"}},"aliases":{},"claims":{"P31":[{"mainsnak":{"snaktype":"value","property":"P31","hash":"fd5912e4dab4b881a8eb0eb27e7893fef55176ad","datavalue":{"value":{"entity-type":"item","numeric-id":56887,"id":"Q56887"},"type":"wikibase-entityid"},"datatype":"wikibase-item"},"type":"statement","id":"Q1983625$81157B16-0624-4CFC-A003-5EF5996965ED","rank":"normal"}],"P159":[{"mainsnak":{"snaktype":"value","property":"P159","hash":"fca9ce754bd6ccd26adfe9b2a83a57cd4b9cae32","datavalue":{"value":{"text":"Over-parametrized deep neural networks minimizing the empirical risk do not generalize well","language":"en"},"type":"monolingualtext"},"datatype":"monolingualtext"},"type":"statement","id":"Q1983625$3AE63C30-D7D7-44D3-9BDC-531F0347C016","rank":"normal"}],"P225":[{"mainsnak":{"snaktype":"value","property":"P225","hash":"be51e1f9cb3ae2a2af2cdf092065487eca06e2ac","datavalue":{"value":"1504.62052","type":"string"},"datatype":"external-id"},"type":"statement","id":"Q1983625$B8AB584A-D583-4E1B-A8B2-BB1A8C0D5CED","rank":"normal"}],"P27":[{"mainsnak":{"snaktype":"value","property":"P27","hash":"981a5c16a3cf3c0a8fb6d8c3f390bacc58a123a2","datavalue":{"value":"10.3150/21-BEJ1323","type":"string"},"datatype":"external-id"},"type":"statement","id":"Q1983625$5083DB88-87DB-4026-B44B-259690EC45E3","rank":"normal"}],"P16":[{"mainsnak":{"snaktype":"value","property":"P16","hash":"306fc02fe8028ee93c41ac57aef8cef85b0efcd6","datavalue":{"value":{"entity-type":"item","numeric-id":225682,"id":"Q225682"},"type":"wikibase-entityid"},"datatype":"wikibase-item"},"type":"statement","id":"Q1983625$AB6E285C-8C51-4E24-B623-E1EEA23C66F8","rank":"normal"},{"mainsnak":{"snaktype":"value","property":"P16","hash":"b9ad97c299022a221b7115ad1e527aaca1dbf0b1","datavalue":{"value":{"entity-type":"item","numeric-id":383852,"id":"Q383852"},"type":"wikibase-entityid"},"datatype":"wikibase-item"},"type":"statement","id":"Q1983625$2514CCE2-F59D-47F3-B869-8E17AC396399","rank":"normal"}],"P200":[{"mainsnak":{"snaktype":"value","property":"P200","hash":"47dbaf2050ccec76d72251c8ab98a49b4485bc38","datavalue":{"value":{"entity-type":"item","numeric-id":61790,"id":"Q61790"},"type":"wikibase-entityid"},"datatype":"wikibase-item"},"type":"statement","id":"Q1983625$D39EDAB0-8974-4757-9250-4EA807BB4A24","rank":"normal"}],"P28":[{"mainsnak":{"snaktype":"value","property":"P28","hash":"b1071fdb9fb6a3f16ad45fa2d11236838d116a1d","datavalue":{"value":{"time":"+2021-09-10T00:00:00Z","timezone":0,"before":0,"after":0,"precision":11,"calendarmodel":"http://www.wikidata.org/entity/Q1985727"},"type":"time"},"datatype":"time"},"type":"statement","id":"Q1983625$24808360-F5E8-4B3D-94B6-70AF46A864BF","rank":"normal"}],"P205":[{"mainsnak":{"snaktype":"value","property":"P205","hash":"87c11c3bcc4f5ebdf327754737763d784c60c53b","datavalue":{"value":"https://arxiv.org/abs/1912.03925","type":"string"},"datatype":"url"},"type":"statement","id":"Q1983625$D1E08C4C-7F55-4F9A-9B56-A973E46E1C84","rank":"normal"}],"P1448":[{"mainsnak":{"snaktype":"value","property":"P1448","hash":"4a97075d382f1ce023952857e1b4d800d8e5a456","datavalue":{"value":"The authors contribute to the theoretical understanding of convergence and generalization properties of neural networks.  They focus on fully connected neural networks with the sigmoidal squasher activation function in a regression setting. The authors find the global minimum of the empirical risk on the training data using over-parametrization. They give a lower bound to achieve a minimal error on the training data with a high probability and prove that such networks do not generalize well on new data. Specifically, they demonstrate how these networks, although minimizing the empirical risk, do not achieve the optimal convergence for estimation of smooth regression functions. Their Theorem 2 shows that any estimates (such as those stated explicitly in Theorem 1) that probabilistically minimize error on the training data do not, in general, generalize well to new data. (They assume that the distributions of \\(X\\) concentrate on finite sets).  The main takeaway from this paper is a somewhat negative result for this kind of fully connected neural network architecture with this type of sigmoidal activation function. The learning process of an over-parametrized neural network does not matter. It cannot reach the optimal minimax convergence rate when it achieves a minimal empirical risk.  In conclusion, it is not clear whether an over-parametrized neural network that minimizes the empirical \\(L_{2}\\) risk generalizes well on new data.","type":"string"},"datatype":"string"},"type":"statement","id":"Q1983625$178F065D-0F28-4208-B9FA-5D592EDD73BC","rank":"normal"}],"P1447":[{"mainsnak":{"snaktype":"value","property":"P1447","hash":"2d5883de915d153e652875f2e6042752b599bb69","datavalue":{"value":{"entity-type":"item","numeric-id":425153,"id":"Q425153"},"type":"wikibase-entityid"},"datatype":"wikibase-item"},"type":"statement","id":"Q1983625$41EE9E9D-1B53-456D-93B4-D7417BE626EA","rank":"normal"}],"P226":[{"mainsnak":{"snaktype":"value","property":"P226","hash":"f246370bc14817a436fbd110d855471daa5fed99","datavalue":{"value":"62G08","type":"string"},"datatype":"external-id"},"type":"statement","id":"Q1983625$381F34D3-8F48-413B-9D58-AC334FCF7ADF","rank":"normal"},{"mainsnak":{"snaktype":"value","property":"P226","hash":"ea3475d4c22dc51a420786874bb708769d5bfd82","datavalue":{"value":"62G20","type":"string"},"datatype":"external-id"},"type":"statement","id":"Q1983625$8F41A2A7-63AE-472B-B4E1-E3633CF3B9FC","rank":"normal"},{"mainsnak":{"snaktype":"value","property":"P226","hash":"f3be836b9b7ef2fd5763308584af4a56d6e109af","datavalue":{"value":"62M45","type":"string"},"datatype":"external-id"},"type":"statement","id":"Q1983625$8D3D67DB-260C-4E20-B563-7382DA268379","rank":"normal"},{"mainsnak":{"snaktype":"value","property":"P226","hash":"259a8687c41a7dbae409f14b2a740ff876886713","datavalue":{"value":"62R07","type":"string"},"datatype":"external-id"},"type":"statement","id":"Q1983625$F8FC03EF-CB4C-4D1A-85EA-3518F89189FD","rank":"normal"},{"mainsnak":{"snaktype":"value","property":"P226","hash":"86ef24f5c9e2aac48660aa2e36054c8246d3fc97","datavalue":{"value":"68T07","type":"string"},"datatype":"external-id"},"type":"statement","id":"Q1983625$95285976-F355-457F-8C6E-BCF4021F6F01","rank":"normal"}],"P1451":[{"mainsnak":{"snaktype":"value","property":"P1451","hash":"de6332f6c4d99aee4a2d37694469c29eee8c2614","datavalue":{"value":"7394101","type":"string"},"datatype":"external-id"},"type":"statement","id":"Q1983625$B3903F93-6650-4013-90E8-5429FDC51FEF","rank":"normal"}],"P1450":[{"mainsnak":{"snaktype":"value","property":"P1450","hash":"b0efe91da4df536f9ec53a9f276b6c2c113d29db","datavalue":{"value":"neural networks","type":"string"},"datatype":"string"},"type":"statement","id":"Q1983625$6BAFA3E9-E5A4-45EF-B008-D4AF41C095E8","rank":"normal"},{"mainsnak":{"snaktype":"value","property":"P1450","hash":"8df4fa885c716a7fd434345dc2599299289d86e0","datavalue":{"value":"nonparametric regression","type":"string"},"datatype":"string"},"type":"statement","id":"Q1983625$E710E0AE-1B56-42A2-BB17-FF5EE5CCD433","rank":"normal"},{"mainsnak":{"snaktype":"value","property":"P1450","hash":"129784650184edbf4854aea45e95340bba6d4c75","datavalue":{"value":"over-parametrization","type":"string"},"datatype":"string"},"type":"statement","id":"Q1983625$944C37ED-83B9-416C-B964-0084447EA6FB","rank":"normal"},{"mainsnak":{"snaktype":"value","property":"P1450","hash":"5097db38d6dca97fdfc6bc38bcb12c7dcfbf08b7","datavalue":{"value":"rate of convergence","type":"string"},"datatype":"string"},"type":"statement","id":"Q1983625$D605DA2F-7269-46AB-82CE-B9AE750834D6","rank":"normal"}],"P1460":[{"mainsnak":{"snaktype":"value","property":"P1460","hash":"57f7fea50d2ce1b39b695c4a1313582eed405e38","datavalue":{"value":{"entity-type":"item","numeric-id":5976449,"id":"Q5976449"},"type":"wikibase-entityid"},"datatype":"wikibase-item"},"type":"statement","id":"Q1983625$6776A1DB-7F84-4413-9456-D9CE63555513","rank":"normal"}],"P223":[{"mainsnak":{"snaktype":"value","property":"P223","hash":"3359f431ed0fabaac556801e32bb9e461edffee6","datavalue":{"value":{"entity-type":"item","numeric-id":5073215,"id":"Q5073215"},"type":"wikibase-entityid"},"datatype":"wikibase-item"},"type":"statement","id":"Q1983625$D580A279-E3D9-4B19-B71D-4F5EAD6148E6","rank":"normal"},{"mainsnak":{"snaktype":"value","property":"P223","hash":"ba791f3bb80cb056c62b0f8a8bcb3a7572a240df","datavalue":{"value":{"entity-type":"item","numeric-id":2313286,"id":"Q2313286"},"type":"wikibase-entityid"},"datatype":"wikibase-item"},"type":"statement","id":"Q1983625$12910DE6-8C7D-4FDD-8FA4-0E1B63CC2376","rank":"normal"},{"mainsnak":{"snaktype":"value","property":"P223","hash":"17a4bcc67e81f7d5c5535afbe003c70c2532ce49","datavalue":{"value":{"entity-type":"item","numeric-id":5218544,"id":"Q5218544"},"type":"wikibase-entityid"},"datatype":"wikibase-item"},"type":"statement","id":"Q1983625$6DDD03BC-6955-4283-A8EC-371143E2859E","rank":"normal"},{"mainsnak":{"snaktype":"value","property":"P223","hash":"f51d92322f978fb96233431f180d05d596e9f355","datavalue":{"value":{"entity-type":"item","numeric-id":4881152,"id":"Q4881152"},"type":"wikibase-entityid"},"datatype":"wikibase-item"},"type":"statement","id":"Q1983625$6EC58AC9-4C67-483E-A2DE-A4A3BEF72366","rank":"normal"},{"mainsnak":{"snaktype":"value","property":"P223","hash":"8679e1051915d59f69a2db9ce8041bfe857a30f4","datavalue":{"value":{"entity-type":"item","numeric-id":1138316,"id":"Q1138316"},"type":"wikibase-entityid"},"datatype":"wikibase-item"},"type":"statement","id":"Q1983625$57359C79-45CF-415E-BADC-7A1309C45EBF","rank":"normal"},{"mainsnak":{"snaktype":"value","property":"P223","hash":"b84ff3422101063f959ec933e950e0ea9ba98695","datavalue":{"value":{"entity-type":"item","numeric-id":1847952,"id":"Q1847952"},"type":"wikibase-entityid"},"datatype":"wikibase-item"},"type":"statement","id":"Q1983625$6BF5A502-4B01-4406-9BCC-A3EC828F755F","rank":"normal"},{"mainsnak":{"snaktype":"value","property":"P223","hash":"ac08e25f9792e2854efe0a51f5d15575885e420e","datavalue":{"value":{"entity-type":"item","numeric-id":5280857,"id":"Q5280857"},"type":"wikibase-entityid"},"datatype":"wikibase-item"},"type":"statement","id":"Q1983625$3EB29DF6-8AA1-48F0-B262-449376DFB895","rank":"normal"},{"mainsnak":{"snaktype":"value","property":"P223","hash":"c03bd2a017a4afc6b6bc41071678b2a1784ec416","datavalue":{"value":{"entity-type":"item","numeric-id":2215715,"id":"Q2215715"},"type":"wikibase-entityid"},"datatype":"wikibase-item"},"type":"statement","id":"Q1983625$29CB2358-EB40-4A2C-A6F8-2ECAD7878E1E","rank":"normal"},{"mainsnak":{"snaktype":"value","property":"P223","hash":"1120e385483b8fa1f92dc55cb2976630f943561c","datavalue":{"value":{"entity-type":"item","numeric-id":2215717,"id":"Q2215717"},"type":"wikibase-entityid"},"datatype":"wikibase-item"},"type":"statement","id":"Q1983625$B2E6A891-A035-40DC-8D0B-1EF339C7A6C7","rank":"normal"},{"mainsnak":{"snaktype":"value","property":"P223","hash":"dade866458d8fb62ddbf49ebfe47d4caeffdd9c4","datavalue":{"value":{"entity-type":"item","numeric-id":1838795,"id":"Q1838795"},"type":"wikibase-entityid"},"datatype":"wikibase-item"},"type":"statement","id":"Q1983625$2CF3FD23-3165-4E58-B311-367E06B3774A","rank":"normal"},{"mainsnak":{"snaktype":"value","property":"P223","hash":"a6fe5ad910c82d8097714e8706a898e31afc5aa5","datavalue":{"value":{"entity-type":"item","numeric-id":1083806,"id":"Q1083806"},"type":"wikibase-entityid"},"datatype":"wikibase-item"},"type":"statement","id":"Q1983625$56A7E692-D816-47EF-86FB-5706B2718822","rank":"normal"},{"mainsnak":{"snaktype":"value","property":"P223","hash":"af307dc8fbf23c002071ae35b09a0a1001e3be21","datavalue":{"value":{"entity-type":"item","numeric-id":1327837,"id":"Q1327837"},"type":"wikibase-entityid"},"datatype":"wikibase-item"},"type":"statement","id":"Q1983625$878878C8-4EEC-4B6D-B0AB-4A93C7211918","rank":"normal"}],"P388":[{"mainsnak":{"snaktype":"value","property":"P388","hash":"6675a2341dc5a30bcdcb28c3c9a22ce81faa6e66","datavalue":{"value":"W3195511373","type":"string"},"datatype":"external-id"},"type":"statement","id":"Q1983625$ABFDE20A-99E0-4F52-B444-234A234C3A5D","rank":"normal"}],"P1643":[{"mainsnak":{"snaktype":"value","property":"P1643","hash":"cc944781253b17fb38dbbb7015efa8ea1cc18135","datavalue":{"value":{"entity-type":"item","numeric-id":5856249,"id":"Q5856249"},"type":"wikibase-entityid"},"datatype":"wikibase-item"},"type":"statement","qualifiers":{"P1659":[{"snaktype":"value","property":"P1659","hash":"cb7eac0f6294196528d72cdf051c4eebbe590352","datavalue":{"value":{"amount":"+0.6245601773262024","unit":"1"},"type":"quantity"},"datatype":"quantity"}],"P1660":[{"snaktype":"value","property":"P1660","hash":"a327a09ea0305e98d5cf33bd4036320e19f2aed0","datavalue":{"value":{"entity-type":"item","numeric-id":6821328,"id":"Q6821328"},"type":"wikibase-entityid"},"datatype":"wikibase-item"}]},"qualifiers-order":["P1659","P1660"],"id":"Q1983625$541020EE-C4B7-47C5-8AA7-501E54F7F6F4","rank":"normal"},{"mainsnak":{"snaktype":"value","property":"P1643","hash":"843f3fefdee048bf612f264decf1f02f42e59599","datavalue":{"value":{"entity-type":"item","numeric-id":2055056,"id":"Q2055056"},"type":"wikibase-entityid"},"datatype":"wikibase-item"},"type":"statement","qualifiers":{"P1659":[{"snaktype":"value","property":"P1659","hash":"975c5dbf37e455a2c805fb2950cd8514937ffe6e","datavalue":{"value":{"amount":"+0.6150848269462585","unit":"1"},"type":"quantity"},"datatype":"quantity"}],"P1660":[{"snaktype":"value","property":"P1660","hash":"a327a09ea0305e98d5cf33bd4036320e19f2aed0","datavalue":{"value":{"entity-type":"item","numeric-id":6821328,"id":"Q6821328"},"type":"wikibase-entityid"},"datatype":"wikibase-item"}]},"qualifiers-order":["P1659","P1660"],"id":"Q1983625$63A5CC0F-A125-49B9-9893-62E9A3102ACE","rank":"normal"},{"mainsnak":{"snaktype":"value","property":"P1643","hash":"05edb684b0050c2d92e38f8ba75e79af9ea7d269","datavalue":{"value":{"entity-type":"item","numeric-id":2034567,"id":"Q2034567"},"type":"wikibase-entityid"},"datatype":"wikibase-item"},"type":"statement","qualifiers":{"P1659":[{"snaktype":"value","property":"P1659","hash":"660b4eb669976fd6f1e4b6c4adadc0a7f68181a9","datavalue":{"value":{"amount":"+0.6148414611816406","unit":"1"},"type":"quantity"},"datatype":"quantity"}],"P1660":[{"snaktype":"value","property":"P1660","hash":"a327a09ea0305e98d5cf33bd4036320e19f2aed0","datavalue":{"value":{"entity-type":"item","numeric-id":6821328,"id":"Q6821328"},"type":"wikibase-entityid"},"datatype":"wikibase-item"}]},"qualifiers-order":["P1659","P1660"],"id":"Q1983625$31D22CEF-4B83-4E8E-A7A1-97B4651178AA","rank":"normal"},{"mainsnak":{"snaktype":"value","property":"P1643","hash":"dacbb2b7fab44bab66e83577ae7568eca14986fb","datavalue":{"value":{"entity-type":"item","numeric-id":3296180,"id":"Q3296180"},"type":"wikibase-entityid"},"datatype":"wikibase-item"},"type":"statement","qualifiers":{"P1659":[{"snaktype":"value","property":"P1659","hash":"ce6b4bd7498bd4025100409367c7851832aca93e","datavalue":{"value":{"amount":"+0.614010751247406","unit":"1"},"type":"quantity"},"datatype":"quantity"}],"P1660":[{"snaktype":"value","property":"P1660","hash":"a327a09ea0305e98d5cf33bd4036320e19f2aed0","datavalue":{"value":{"entity-type":"item","numeric-id":6821328,"id":"Q6821328"},"type":"wikibase-entityid"},"datatype":"wikibase-item"}]},"qualifiers-order":["P1659","P1660"],"id":"Q1983625$1A6FE699-E70C-4DB4-BB83-761436C64B68","rank":"normal"},{"mainsnak":{"snaktype":"value","property":"P1643","hash":"309a0e29a6693be026b897c388c49e1175abcab1","datavalue":{"value":{"entity-type":"item","numeric-id":2185668,"id":"Q2185668"},"type":"wikibase-entityid"},"datatype":"wikibase-item"},"type":"statement","qualifiers":{"P1659":[{"snaktype":"value","property":"P1659","hash":"9eec92db187e1d633671b902b31138f3496fc0a3","datavalue":{"value":{"amount":"+0.6121954321861267","unit":"1"},"type":"quantity"},"datatype":"quantity"}],"P1660":[{"snaktype":"value","property":"P1660","hash":"a327a09ea0305e98d5cf33bd4036320e19f2aed0","datavalue":{"value":{"entity-type":"item","numeric-id":6821328,"id":"Q6821328"},"type":"wikibase-entityid"},"datatype":"wikibase-item"}]},"qualifiers-order":["P1659","P1660"],"id":"Q1983625$96BC8379-D9DC-447E-84AF-1E454DF8EA13","rank":"normal"}]},"sitelinks":{"mardi":{"site":"mardi","title":"Over-parametrized deep neural networks minimizing the empirical risk do not generalize well","badges":[],"url":"https://portal.mardi4nfdi.de/wiki/Over-parametrized_deep_neural_networks_minimizing_the_empirical_risk_do_not_generalize_well"}}}}}