Matches in SemOpenAlex for { <https://semopenalex.org/work/W2898859254> ?p ?o ?g. }
- W2898859254 abstract "We study the loss surface of a feed-forward neural network with ReLU non-linearities, regularized with weight decay. We show that the regularized loss function is piecewise strongly convex on an important open set which contains, under some conditions, all of its global minimizers. This is used to prove that local minima of the regularized loss function in this set are isolated, and that every differentiable critical point in this set is a local minimum, partially addressing an open problem given at the Conference on Learning Theory (COLT) 2015; our result is also applied to linear neural networks to show that with weight decay regularization, there are no non-zero critical points in a norm ball obtaining training error below a given threshold. We also include an experimental section where we validate our theoretical work and show that the regularized loss function is almost always piecewise strongly convex when restricted to stochastic gradient descent trajectories for three standard image classification problems." @default.
- W2898859254 created "2018-11-09" @default.
- W2898859254 creator A5045592643 @default.
- W2898859254 date "2018-10-30" @default.
- W2898859254 modified "2023-09-27" @default.
- W2898859254 title "Piecewise Strong Convexity of Neural Networks." @default.
- W2898859254 cites W1899249567 @default.
- W2898859254 cites W1964694549 @default.
- W2898859254 cites W1975841865 @default.
- W2898859254 cites W2095705004 @default.
- W2898859254 cites W2112796928 @default.
- W2898859254 cites W2125930537 @default.
- W2898859254 cites W2215006509 @default.
- W2898859254 cites W2399994860 @default.
- W2898859254 cites W2474090883 @default.
- W2898859254 cites W2601424732 @default.
- W2898859254 cites W2894842463 @default.
- W2898859254 cites W2899748887 @default.
- W2898859254 cites W2913892099 @default.
- W2898859254 cites W294519166 @default.
- W2898859254 cites W2949650786 @default.
- W2898859254 cites W2950220847 @default.
- W2898859254 cites W2952574409 @default.
- W2898859254 cites W2963326517 @default.
- W2898859254 cites W2963399222 @default.
- W2898859254 cites W2963446085 @default.
- W2898859254 cites W2963800363 @default.
- W2898859254 hasPublicationYear "2018" @default.
- W2898859254 type Work @default.
- W2898859254 sameAs 2898859254 @default.
- W2898859254 citedByCount "1" @default.
- W2898859254 countsByYear W28988592542019 @default.
- W2898859254 crossrefType "posted-content" @default.
- W2898859254 hasAuthorship W2898859254A5045592643 @default.
- W2898859254 hasConcept C106159729 @default.
- W2898859254 hasConcept C112680207 @default.
- W2898859254 hasConcept C126255220 @default.
- W2898859254 hasConcept C134306372 @default.
- W2898859254 hasConcept C145446738 @default.
- W2898859254 hasConcept C153258448 @default.
- W2898859254 hasConcept C154945302 @default.
- W2898859254 hasConcept C162324750 @default.
- W2898859254 hasConcept C164660894 @default.
- W2898859254 hasConcept C17744445 @default.
- W2898859254 hasConcept C186633575 @default.
- W2898859254 hasConcept C191795146 @default.
- W2898859254 hasConcept C199539241 @default.
- W2898859254 hasConcept C202444582 @default.
- W2898859254 hasConcept C202615002 @default.
- W2898859254 hasConcept C206688291 @default.
- W2898859254 hasConcept C2524010 @default.
- W2898859254 hasConcept C2776135515 @default.
- W2898859254 hasConcept C28826006 @default.
- W2898859254 hasConcept C33923547 @default.
- W2898859254 hasConcept C41008148 @default.
- W2898859254 hasConcept C42357961 @default.
- W2898859254 hasConcept C50644808 @default.
- W2898859254 hasConcept C72134830 @default.
- W2898859254 hasConceptScore W2898859254C106159729 @default.
- W2898859254 hasConceptScore W2898859254C112680207 @default.
- W2898859254 hasConceptScore W2898859254C126255220 @default.
- W2898859254 hasConceptScore W2898859254C134306372 @default.
- W2898859254 hasConceptScore W2898859254C145446738 @default.
- W2898859254 hasConceptScore W2898859254C153258448 @default.
- W2898859254 hasConceptScore W2898859254C154945302 @default.
- W2898859254 hasConceptScore W2898859254C162324750 @default.
- W2898859254 hasConceptScore W2898859254C164660894 @default.
- W2898859254 hasConceptScore W2898859254C17744445 @default.
- W2898859254 hasConceptScore W2898859254C186633575 @default.
- W2898859254 hasConceptScore W2898859254C191795146 @default.
- W2898859254 hasConceptScore W2898859254C199539241 @default.
- W2898859254 hasConceptScore W2898859254C202444582 @default.
- W2898859254 hasConceptScore W2898859254C202615002 @default.
- W2898859254 hasConceptScore W2898859254C206688291 @default.
- W2898859254 hasConceptScore W2898859254C2524010 @default.
- W2898859254 hasConceptScore W2898859254C2776135515 @default.
- W2898859254 hasConceptScore W2898859254C28826006 @default.
- W2898859254 hasConceptScore W2898859254C33923547 @default.
- W2898859254 hasConceptScore W2898859254C41008148 @default.
- W2898859254 hasConceptScore W2898859254C42357961 @default.
- W2898859254 hasConceptScore W2898859254C50644808 @default.
- W2898859254 hasConceptScore W2898859254C72134830 @default.
- W2898859254 hasLocation W28988592541 @default.
- W2898859254 hasOpenAccess W2898859254 @default.
- W2898859254 hasPrimaryLocation W28988592541 @default.
- W2898859254 hasRelatedWork W1179775547 @default.
- W2898859254 hasRelatedWork W136389534 @default.
- W2898859254 hasRelatedWork W1697075315 @default.
- W2898859254 hasRelatedWork W2034645553 @default.
- W2898859254 hasRelatedWork W2113564814 @default.
- W2898859254 hasRelatedWork W2114609025 @default.
- W2898859254 hasRelatedWork W2198227614 @default.
- W2898859254 hasRelatedWork W2272509765 @default.
- W2898859254 hasRelatedWork W2647383441 @default.
- W2898859254 hasRelatedWork W2790417304 @default.
- W2898859254 hasRelatedWork W2797791799 @default.
- W2898859254 hasRelatedWork W2811361967 @default.
- W2898859254 hasRelatedWork W2893172655 @default.
- W2898859254 hasRelatedWork W3005835621 @default.
- W2898859254 hasRelatedWork W3006873381 @default.