FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3132829052> ?p ?o ?g. }

• W3132829052 abstract "We describe the convex semi-infinite dual of the two-layer vector-output ReLU neural network training problem. This semi-infinite dual admits a finite dimensional representation, but its support is over a convex set which is difficult to characterize. In particular, we demonstrate that the non-convex neural network training problem is equivalent to a finite-dimensional convex copositive program. Our work is the first to identify this strong connection between the global optima of neural networks and those of copositive programs. We thus demonstrate how neural networks implicitly attempt to solve copositive programs via semi-nonnegative matrix factorization, and draw key insights from this formulation. We describe the first algorithms for provably finding the global minimum of the vector output neural network training problem, which are polynomial in the number of samples for a fixed data rank, yet exponential in the dimension. However, in the case of convolutional architectures, the computational complexity is exponential in only the filter size and polynomial in all other parameters. We describe the circumstances in which we can find the global optimum of this neural network training problem exactly with soft-thresholded SVD, and provide a copositive relaxation which is guaranteed to be exact for certain classes of problems, and which corresponds with the solution of Stochastic Gradient Descent in practice." @default.
• W3132829052 created "2021-03-01" @default.
• W3132829052 creator A5001436196 @default.
• W3132829052 creator A5008348052 @default.
• W3132829052 creator A5040173784 @default.
• W3132829052 creator A5079543660 @default.
• W3132829052 date "2021-05-03" @default.
• W3132829052 modified "2023-09-24" @default.
• W3132829052 title "Vector-output ReLU Neural Network Problems are Copositive Programs: Convex Analysis of Two Layer Networks and Polynomial-time Algorithms" @default.
• W3132829052 cites W1509803206 @default.
• W3132829052 cites W1677182931 @default.
• W3132829052 cites W1972746552 @default.
• W3132829052 cites W2043136268 @default.
• W3132829052 cites W2065180801 @default.
• W3132829052 cites W2088536697 @default.
• W3132829052 cites W2110531331 @default.
• W3132829052 cites W2112796928 @default.
• W3132829052 cites W2116413942 @default.
• W3132829052 cites W2118550318 @default.
• W3132829052 cites W2122090912 @default.
• W3132829052 cites W2124487034 @default.
• W3132829052 cites W2127553237 @default.
• W3132829052 cites W2131637439 @default.
• W3132829052 cites W2134332047 @default.
• W3132829052 cites W2136885855 @default.
• W3132829052 cites W2168103112 @default.
• W3132829052 cites W2504734010 @default.
• W3132829052 cites W2606477047 @default.
• W3132829052 cites W2608666998 @default.
• W3132829052 cites W2803183915 @default.
• W3132829052 cites W2809090039 @default.
• W3132829052 cites W2899476926 @default.
• W3132829052 cites W2914984858 @default.
• W3132829052 cites W2946790101 @default.
• W3132829052 cites W2949798199 @default.
• W3132829052 cites W2952318479 @default.
• W3132829052 cites W2963047948 @default.
• W3132829052 cites W2963695615 @default.
• W3132829052 cites W2963743626 @default.
• W3132829052 cites W2964121744 @default.
• W3132829052 cites W2970176896 @default.
• W3132829052 cites W2971043187 @default.
• W3132829052 cites W3006926186 @default.
• W3132829052 cites W3034552778 @default.
• W3132829052 cites W3037536384 @default.
• W3132829052 cites W3118608800 @default.
• W3132829052 cites W37129894 @default.
• W3132829052 cites W1962325108 @default.
• W3132829052 cites W2290452516 @default.
• W3132829052 hasPublicationYear "2021" @default.
• W3132829052 type Work @default.
• W3132829052 sameAs 3132829052 @default.
• W3132829052 citedByCount "10" @default.
• W3132829052 countsByYear W31328290522020 @default.
• W3132829052 countsByYear W31328290522021 @default.
• W3132829052 crossrefType "proceedings-article" @default.
• W3132829052 hasAuthorship W3132829052A5001436196 @default.
• W3132829052 hasAuthorship W3132829052A5008348052 @default.
• W3132829052 hasAuthorship W3132829052A5040173784 @default.
• W3132829052 hasAuthorship W3132829052A5079543660 @default.
• W3132829052 hasConcept C112680207 @default.
• W3132829052 hasConcept C11413529 @default.
• W3132829052 hasConcept C126255220 @default.
• W3132829052 hasConcept C134306372 @default.
• W3132829052 hasConcept C154945302 @default.
• W3132829052 hasConcept C15744967 @default.
• W3132829052 hasConcept C157972887 @default.
• W3132829052 hasConcept C2524010 @default.
• W3132829052 hasConcept C2776029896 @default.
• W3132829052 hasConcept C33923547 @default.
• W3132829052 hasConcept C41008148 @default.
• W3132829052 hasConcept C50644808 @default.
• W3132829052 hasConcept C77805123 @default.
• W3132829052 hasConcept C90119067 @default.
• W3132829052 hasConceptScore W3132829052C112680207 @default.
• W3132829052 hasConceptScore W3132829052C11413529 @default.
• W3132829052 hasConceptScore W3132829052C126255220 @default.
• W3132829052 hasConceptScore W3132829052C134306372 @default.
• W3132829052 hasConceptScore W3132829052C154945302 @default.
• W3132829052 hasConceptScore W3132829052C15744967 @default.
• W3132829052 hasConceptScore W3132829052C157972887 @default.
• W3132829052 hasConceptScore W3132829052C2524010 @default.
• W3132829052 hasConceptScore W3132829052C2776029896 @default.
• W3132829052 hasConceptScore W3132829052C33923547 @default.
• W3132829052 hasConceptScore W3132829052C41008148 @default.
• W3132829052 hasConceptScore W3132829052C50644808 @default.
• W3132829052 hasConceptScore W3132829052C77805123 @default.
• W3132829052 hasConceptScore W3132829052C90119067 @default.
• W3132829052 hasLocation W31328290521 @default.
• W3132829052 hasOpenAccess W3132829052 @default.
• W3132829052 hasPrimaryLocation W31328290521 @default.
• W3132829052 hasRelatedWork W1509722456 @default.
• W3132829052 hasRelatedWork W2025762001 @default.
• W3132829052 hasRelatedWork W2032637534 @default.
• W3132829052 hasRelatedWork W2061309766 @default.
• W3132829052 hasRelatedWork W2097139389 @default.
• W3132829052 hasRelatedWork W2161278885 @default.
• W3132829052 hasRelatedWork W2573095247 @default.
• W3132829052 hasRelatedWork W2608666998 @default.
• W3132829052 hasRelatedWork W2790814024 @default.

• next