Matches in SemOpenAlex for { <https://semopenalex.org/work/W3103322710> ?p ?o ?g. }
- W3103322710 abstract "Recent work linking deep neural networks and dynamical systems opened up new avenues to analyze deep learning. In particular, it is observed that new insights can be obtained by recasting deep learning as an optimal control problem on difference or differential equations. However, the mathematical aspects of such a formulation have not been systematically explored. This paper introduces the mathematical formulation of the population risk minimization problem in deep learning as a mean-field optimal control problem. Mirroring the development of classical optimal control, we state and prove optimality conditions of both the Hamilton-Jacobi-Bellman type and the Pontryagin type. These mean-field results reflect the probabilistic nature of the learning problem. In addition, by appealing to the mean-field Pontryagin's maximum principle, we establish some quantitative relationships between population and empirical learning problems. This serves to establish a mathematical foundation for investigating the algorithmic and theoretical connections between optimal control and deep learning." @default.
- W3103322710 created "2020-11-23" @default.
- W3103322710 creator A5017947325 @default.
- W3103322710 creator A5069654038 @default.
- W3103322710 creator A5071854504 @default.
- W3103322710 date "2018-12-13" @default.
- W3103322710 modified "2023-10-16" @default.
- W3103322710 title "A mean-field optimal control formulation of deep learning" @default.
- W3103322710 cites W1637901183 @default.
- W3103322710 cites W1820663283 @default.
- W3103322710 cites W182704713 @default.
- W3103322710 cites W1969717343 @default.
- W3103322710 cites W1979308690 @default.
- W3103322710 cites W2011000015 @default.
- W3103322710 cites W2024287949 @default.
- W3103322710 cites W2032316144 @default.
- W3103322710 cites W2035845388 @default.
- W3103322710 cites W2037152246 @default.
- W3103322710 cites W2038686546 @default.
- W3103322710 cites W2040800769 @default.
- W3103322710 cites W2045400766 @default.
- W3103322710 cites W2053918727 @default.
- W3103322710 cites W2065908952 @default.
- W3103322710 cites W2115733720 @default.
- W3103322710 cites W2131127275 @default.
- W3103322710 cites W2148316915 @default.
- W3103322710 cites W2165698076 @default.
- W3103322710 cites W2194775991 @default.
- W3103322710 cites W2298991554 @default.
- W3103322710 cites W2493294075 @default.
- W3103322710 cites W2600297185 @default.
- W3103322710 cites W2749408143 @default.
- W3103322710 cites W2919115771 @default.
- W3103322710 cites W2952892930 @default.
- W3103322710 cites W2962818191 @default.
- W3103322710 cites W2963359731 @default.
- W3103322710 cites W2963415493 @default.
- W3103322710 cites W3098011980 @default.
- W3103322710 cites W4231109964 @default.
- W3103322710 cites W4301747304 @default.
- W3103322710 cites W586490843 @default.
- W3103322710 doi "https://doi.org/10.1007/s40687-018-0172-y" @default.
- W3103322710 hasPublicationYear "2018" @default.
- W3103322710 type Work @default.
- W3103322710 sameAs 3103322710 @default.
- W3103322710 citedByCount "93" @default.
- W3103322710 countsByYear W31033227102019 @default.
- W3103322710 countsByYear W31033227102020 @default.
- W3103322710 countsByYear W31033227102021 @default.
- W3103322710 countsByYear W31033227102022 @default.
- W3103322710 countsByYear W31033227102023 @default.
- W3103322710 crossrefType "journal-article" @default.
- W3103322710 hasAuthorship W3103322710A5017947325 @default.
- W3103322710 hasAuthorship W3103322710A5069654038 @default.
- W3103322710 hasAuthorship W3103322710A5071854504 @default.
- W3103322710 hasBestOaLocation W31033227102 @default.
- W3103322710 hasConcept C108583219 @default.
- W3103322710 hasConcept C126255220 @default.
- W3103322710 hasConcept C144024400 @default.
- W3103322710 hasConcept C149923435 @default.
- W3103322710 hasConcept C154945302 @default.
- W3103322710 hasConcept C202444582 @default.
- W3103322710 hasConcept C2908647359 @default.
- W3103322710 hasConcept C33923547 @default.
- W3103322710 hasConcept C41008148 @default.
- W3103322710 hasConcept C91575142 @default.
- W3103322710 hasConcept C9652623 @default.
- W3103322710 hasConceptScore W3103322710C108583219 @default.
- W3103322710 hasConceptScore W3103322710C126255220 @default.
- W3103322710 hasConceptScore W3103322710C144024400 @default.
- W3103322710 hasConceptScore W3103322710C149923435 @default.
- W3103322710 hasConceptScore W3103322710C154945302 @default.
- W3103322710 hasConceptScore W3103322710C202444582 @default.
- W3103322710 hasConceptScore W3103322710C2908647359 @default.
- W3103322710 hasConceptScore W3103322710C33923547 @default.
- W3103322710 hasConceptScore W3103322710C41008148 @default.
- W3103322710 hasConceptScore W3103322710C91575142 @default.
- W3103322710 hasConceptScore W3103322710C9652623 @default.
- W3103322710 hasFunder F4320321001 @default.
- W3103322710 hasFunder F4320337345 @default.
- W3103322710 hasIssue "1" @default.
- W3103322710 hasLocation W31033227101 @default.
- W3103322710 hasLocation W31033227102 @default.
- W3103322710 hasLocation W31033227103 @default.
- W3103322710 hasOpenAccess W3103322710 @default.
- W3103322710 hasPrimaryLocation W31033227101 @default.
- W3103322710 hasRelatedWork W2611989081 @default.
- W3103322710 hasRelatedWork W2731899572 @default.
- W3103322710 hasRelatedWork W2802049774 @default.
- W3103322710 hasRelatedWork W3215138031 @default.
- W3103322710 hasRelatedWork W4230611425 @default.
- W3103322710 hasRelatedWork W4294635752 @default.
- W3103322710 hasRelatedWork W4304166257 @default.
- W3103322710 hasRelatedWork W4327774331 @default.
- W3103322710 hasRelatedWork W4375867731 @default.
- W3103322710 hasRelatedWork W4383066092 @default.
- W3103322710 hasVolume "6" @default.
- W3103322710 isParatext "false" @default.
- W3103322710 isRetracted "false" @default.
- W3103322710 magId "3103322710" @default.