Matches in SemOpenAlex for { <https://semopenalex.org/work/W2901816197> ?p ?o ?g. }
- W2901816197 endingPage "39" @default.
- W2901816197 startingPage "3" @default.
- W2901816197 abstract "We present strong mixed-integer programming (MIP) formulations for high-dimensional piecewise linear functions that correspond to trained neural networks. These formulations can be used for a number of important tasks, such as verifying that an image classification network is robust to adversarial inputs, or solving decision problems where the objective function is a machine learning model. We present a generic framework, which may be of independent interest, that provides a way to construct sharp or ideal formulations for the maximum of d affine functions over arbitrary polyhedral input domains. We apply this result to derive MIP formulations for a number of the most popular nonlinear operations (e.g. ReLU and max pooling) that are strictly stronger than other approaches from the literature. We corroborate this computationally, showing that our formulations are able to offer substantial improvements in solve time on verification tasks for image classification networks." @default.
- W2901816197 created "2018-11-29" @default.
- W2901816197 creator A5008961628 @default.
- W2901816197 creator A5021227744 @default.
- W2901816197 creator A5039507401 @default.
- W2901816197 creator A5048414452 @default.
- W2901816197 creator A5070149920 @default.
- W2901816197 date "2020-02-13" @default.
- W2901816197 modified "2023-10-05" @default.
- W2901816197 title "Strong mixed-integer programming formulations for trained neural networks" @default.
- W2901816197 cites W1019830208 @default.
- W2901816197 cites W10227180 @default.
- W2901816197 cites W127277165 @default.
- W2901816197 cites W155963807 @default.
- W2901816197 cites W1965897919 @default.
- W2901816197 cites W1967284207 @default.
- W2901816197 cites W1972997295 @default.
- W2901816197 cites W1974401856 @default.
- W2901816197 cites W2004548680 @default.
- W2901816197 cites W2027627501 @default.
- W2901816197 cites W2036898642 @default.
- W2901816197 cites W205043357 @default.
- W2901816197 cites W2066071727 @default.
- W2901816197 cites W2073576149 @default.
- W2901816197 cites W2090042335 @default.
- W2901816197 cites W2098425563 @default.
- W2901816197 cites W2105527809 @default.
- W2901816197 cites W2112796928 @default.
- W2901816197 cites W2151308681 @default.
- W2901816197 cites W2180612164 @default.
- W2901816197 cites W2232834753 @default.
- W2901816197 cites W2276412021 @default.
- W2901816197 cites W2345492474 @default.
- W2901816197 cites W2433743436 @default.
- W2901816197 cites W2594877703 @default.
- W2901816197 cites W2741534749 @default.
- W2901816197 cites W2742154030 @default.
- W2901816197 cites W2791251367 @default.
- W2901816197 cites W2797572905 @default.
- W2901816197 cites W2802557767 @default.
- W2901816197 cites W2846823501 @default.
- W2901816197 cites W2908984973 @default.
- W2901816197 cites W2919115771 @default.
- W2901816197 cites W2963054787 @default.
- W2901816197 cites W2963491246 @default.
- W2901816197 cites W2963600714 @default.
- W2901816197 cites W2963857521 @default.
- W2901816197 cites W2976145414 @default.
- W2901816197 cites W2998757566 @default.
- W2901816197 cites W3100789280 @default.
- W2901816197 cites W3103957401 @default.
- W2901816197 cites W4379506771 @default.
- W2901816197 doi "https://doi.org/10.1007/s10107-020-01474-5" @default.
- W2901816197 hasPublicationYear "2020" @default.
- W2901816197 type Work @default.
- W2901816197 sameAs 2901816197 @default.
- W2901816197 citedByCount "59" @default.
- W2901816197 countsByYear W29018161972017 @default.
- W2901816197 countsByYear W29018161972018 @default.
- W2901816197 countsByYear W29018161972019 @default.
- W2901816197 countsByYear W29018161972020 @default.
- W2901816197 countsByYear W29018161972021 @default.
- W2901816197 countsByYear W29018161972022 @default.
- W2901816197 countsByYear W29018161972023 @default.
- W2901816197 crossrefType "journal-article" @default.
- W2901816197 hasAuthorship W2901816197A5008961628 @default.
- W2901816197 hasAuthorship W2901816197A5021227744 @default.
- W2901816197 hasAuthorship W2901816197A5039507401 @default.
- W2901816197 hasAuthorship W2901816197A5048414452 @default.
- W2901816197 hasAuthorship W2901816197A5070149920 @default.
- W2901816197 hasBestOaLocation W29018161972 @default.
- W2901816197 hasConcept C11413529 @default.
- W2901816197 hasConcept C126255220 @default.
- W2901816197 hasConcept C134306372 @default.
- W2901816197 hasConcept C14036430 @default.
- W2901816197 hasConcept C154945302 @default.
- W2901816197 hasConcept C164660894 @default.
- W2901816197 hasConcept C17095337 @default.
- W2901816197 hasConcept C199360897 @default.
- W2901816197 hasConcept C202444582 @default.
- W2901816197 hasConcept C2524010 @default.
- W2901816197 hasConcept C2780801425 @default.
- W2901816197 hasConcept C33923547 @default.
- W2901816197 hasConcept C41008148 @default.
- W2901816197 hasConcept C41045048 @default.
- W2901816197 hasConcept C50644808 @default.
- W2901816197 hasConcept C56086750 @default.
- W2901816197 hasConcept C78458016 @default.
- W2901816197 hasConcept C86803240 @default.
- W2901816197 hasConcept C92757383 @default.
- W2901816197 hasConcept C97137487 @default.
- W2901816197 hasConceptScore W2901816197C11413529 @default.
- W2901816197 hasConceptScore W2901816197C126255220 @default.
- W2901816197 hasConceptScore W2901816197C134306372 @default.
- W2901816197 hasConceptScore W2901816197C14036430 @default.
- W2901816197 hasConceptScore W2901816197C154945302 @default.
- W2901816197 hasConceptScore W2901816197C164660894 @default.
- W2901816197 hasConceptScore W2901816197C17095337 @default.