Matches in SemOpenAlex for { <https://semopenalex.org/work/W2896083421> ?p ?o ?g. }
Showing items 1 to 78 of
78
with 100 items per page.
- W2896083421 abstract "By using the viewpoint of modern computational algebraic geometry, we explore properties of the optimization landscapes of the deep linear neural network models. After clarifying on the various definitions of minima, we show that the geometrically flat minima, which are merely artifacts of residual continuous symmetries of the deep linear networks, can be straightforwardly removed by a generalized $L_2$ regularization. Then, we establish upper bounds on the number of isolated stationary points of these networks with the help of algebraic geometry. Using these upper bounds and utilizing a numerical algebraic geometry method, we find all stationary points of modest depth and matrix size. We show that in the presence of the non-zero regularization, deep linear networks indeed possess local minima which are not the global minima. Our computational results clarify certain aspects of the loss surfaces of deep linear networks and provide novel insights." @default.
- W2896083421 created "2018-10-26" @default.
- W2896083421 creator A5007214938 @default.
- W2896083421 creator A5034617367 @default.
- W2896083421 creator A5056701307 @default.
- W2896083421 creator A5069434639 @default.
- W2896083421 date "2018-10-17" @default.
- W2896083421 modified "2023-09-23" @default.
- W2896083421 title "The loss surface of deep linear networks viewed through the algebraic geometry lens" @default.
- W2896083421 hasPublicationYear "2018" @default.
- W2896083421 type Work @default.
- W2896083421 sameAs 2896083421 @default.
- W2896083421 citedByCount "7" @default.
- W2896083421 countsByYear W28960834212019 @default.
- W2896083421 countsByYear W28960834212020 @default.
- W2896083421 countsByYear W28960834212021 @default.
- W2896083421 crossrefType "posted-content" @default.
- W2896083421 hasAuthorship W2896083421A5007214938 @default.
- W2896083421 hasAuthorship W2896083421A5034617367 @default.
- W2896083421 hasAuthorship W2896083421A5056701307 @default.
- W2896083421 hasAuthorship W2896083421A5069434639 @default.
- W2896083421 hasConcept C108583219 @default.
- W2896083421 hasConcept C134306372 @default.
- W2896083421 hasConcept C154945302 @default.
- W2896083421 hasConcept C186633575 @default.
- W2896083421 hasConcept C189237950 @default.
- W2896083421 hasConcept C2524010 @default.
- W2896083421 hasConcept C2776135515 @default.
- W2896083421 hasConcept C2776799497 @default.
- W2896083421 hasConcept C33923547 @default.
- W2896083421 hasConcept C41008148 @default.
- W2896083421 hasConcept C50644808 @default.
- W2896083421 hasConcept C68363185 @default.
- W2896083421 hasConcept C69653121 @default.
- W2896083421 hasConcept C9376300 @default.
- W2896083421 hasConcept C96469262 @default.
- W2896083421 hasConceptScore W2896083421C108583219 @default.
- W2896083421 hasConceptScore W2896083421C134306372 @default.
- W2896083421 hasConceptScore W2896083421C154945302 @default.
- W2896083421 hasConceptScore W2896083421C186633575 @default.
- W2896083421 hasConceptScore W2896083421C189237950 @default.
- W2896083421 hasConceptScore W2896083421C2524010 @default.
- W2896083421 hasConceptScore W2896083421C2776135515 @default.
- W2896083421 hasConceptScore W2896083421C2776799497 @default.
- W2896083421 hasConceptScore W2896083421C33923547 @default.
- W2896083421 hasConceptScore W2896083421C41008148 @default.
- W2896083421 hasConceptScore W2896083421C50644808 @default.
- W2896083421 hasConceptScore W2896083421C68363185 @default.
- W2896083421 hasConceptScore W2896083421C69653121 @default.
- W2896083421 hasConceptScore W2896083421C9376300 @default.
- W2896083421 hasConceptScore W2896083421C96469262 @default.
- W2896083421 hasLocation W28960834211 @default.
- W2896083421 hasOpenAccess W2896083421 @default.
- W2896083421 hasPrimaryLocation W28960834211 @default.
- W2896083421 hasRelatedWork W123964978 @default.
- W2896083421 hasRelatedWork W1502898252 @default.
- W2896083421 hasRelatedWork W1770499924 @default.
- W2896083421 hasRelatedWork W1981378358 @default.
- W2896083421 hasRelatedWork W1982356071 @default.
- W2896083421 hasRelatedWork W1986970331 @default.
- W2896083421 hasRelatedWork W1999298886 @default.
- W2896083421 hasRelatedWork W2078626246 @default.
- W2896083421 hasRelatedWork W2118218373 @default.
- W2896083421 hasRelatedWork W2183833388 @default.
- W2896083421 hasRelatedWork W2755882036 @default.
- W2896083421 hasRelatedWork W2910076263 @default.
- W2896083421 hasRelatedWork W2949987090 @default.
- W2896083421 hasRelatedWork W2952103892 @default.
- W2896083421 hasRelatedWork W2952122845 @default.
- W2896083421 hasRelatedWork W2956434516 @default.
- W2896083421 hasRelatedWork W2963720369 @default.
- W2896083421 hasRelatedWork W2972836004 @default.
- W2896083421 hasRelatedWork W3180560770 @default.
- W2896083421 hasRelatedWork W3185200785 @default.
- W2896083421 isParatext "false" @default.
- W2896083421 isRetracted "false" @default.
- W2896083421 magId "2896083421" @default.
- W2896083421 workType "article" @default.