FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2797798064> ?p ?o ?g. }

• W2797798064 abstract "Effective regularisation of neural networks is essential to combat overfitting due to the large number of parameters involved. We present an empirical analogue to the Lipschitz constant of a feed-forward neural network, which we refer to as the maximum gain. We hypothesise that constraining the gain of a network will have a regularising effect, similar to how constraining the Lipschitz constant of a network has been shown to improve generalisation. A simple algorithm is provided that involves rescaling the weight matrix of each layer after each parameter update. We conduct a series of studies on common benchmark datasets, and also a novel dataset that we introduce to enable easier significance testing for experiments using convolutional networks. Performance on these datasets compares favourably with other common regularisation techniques." @default.
• W2797798064 created "2018-04-24" @default.
• W2797798064 creator A5039741908 @default.
• W2797798064 creator A5059992863 @default.
• W2797798064 creator A5087785022 @default.
• W2797798064 creator A5090758218 @default.
• W2797798064 date "2018-04-16" @default.
• W2797798064 modified "2023-09-27" @default.
• W2797798064 title "MaxGain: Regularisation of Neural Networks by Constraining Activation Magnitudes" @default.
• W2797798064 cites W1686810756 @default.
• W2797798064 cites W2095705004 @default.
• W2797798064 cites W2099471712 @default.
• W2797798064 cites W2108598243 @default.
• W2797798064 cites W2401231614 @default.
• W2797798064 cites W2796892552 @default.
• W2797798064 cites W2949117887 @default.
• W2797798064 cites W2950928354 @default.
• W2797798064 cites W2951595529 @default.
• W2797798064 cites W2962879692 @default.
• W2797798064 cites W2963836885 @default.
• W2797798064 cites W2964121744 @default.
• W2797798064 cites W3037590790 @default.
• W2797798064 cites W3118608800 @default.
• W2797798064 cites W3123294050 @default.
• W2797798064 cites W3137695714 @default.
• W2797798064 hasPublicationYear "2018" @default.
• W2797798064 type Work @default.
• W2797798064 sameAs 2797798064 @default.
• W2797798064 citedByCount "0" @default.
• W2797798064 crossrefType "posted-content" @default.
• W2797798064 hasAuthorship W2797798064A5039741908 @default.
• W2797798064 hasAuthorship W2797798064A5059992863 @default.
• W2797798064 hasAuthorship W2797798064A5087785022 @default.
• W2797798064 hasAuthorship W2797798064A5090758218 @default.
• W2797798064 hasConcept C111472728 @default.
• W2797798064 hasConcept C119857082 @default.
• W2797798064 hasConcept C127413603 @default.
• W2797798064 hasConcept C13280743 @default.
• W2797798064 hasConcept C134306372 @default.
• W2797798064 hasConcept C138885662 @default.
• W2797798064 hasConcept C154945302 @default.
• W2797798064 hasConcept C185798385 @default.
• W2797798064 hasConcept C199360897 @default.
• W2797798064 hasConcept C205649164 @default.
• W2797798064 hasConcept C21200559 @default.
• W2797798064 hasConcept C22019652 @default.
• W2797798064 hasConcept C22324862 @default.
• W2797798064 hasConcept C24326235 @default.
• W2797798064 hasConcept C2777027219 @default.
• W2797798064 hasConcept C2780586882 @default.
• W2797798064 hasConcept C33923547 @default.
• W2797798064 hasConcept C41008148 @default.
• W2797798064 hasConcept C50644808 @default.
• W2797798064 hasConcept C81363708 @default.
• W2797798064 hasConceptScore W2797798064C111472728 @default.
• W2797798064 hasConceptScore W2797798064C119857082 @default.
• W2797798064 hasConceptScore W2797798064C127413603 @default.
• W2797798064 hasConceptScore W2797798064C13280743 @default.
• W2797798064 hasConceptScore W2797798064C134306372 @default.
• W2797798064 hasConceptScore W2797798064C138885662 @default.
• W2797798064 hasConceptScore W2797798064C154945302 @default.
• W2797798064 hasConceptScore W2797798064C185798385 @default.
• W2797798064 hasConceptScore W2797798064C199360897 @default.
• W2797798064 hasConceptScore W2797798064C205649164 @default.
• W2797798064 hasConceptScore W2797798064C21200559 @default.
• W2797798064 hasConceptScore W2797798064C22019652 @default.
• W2797798064 hasConceptScore W2797798064C22324862 @default.
• W2797798064 hasConceptScore W2797798064C24326235 @default.
• W2797798064 hasConceptScore W2797798064C2777027219 @default.
• W2797798064 hasConceptScore W2797798064C2780586882 @default.
• W2797798064 hasConceptScore W2797798064C33923547 @default.
• W2797798064 hasConceptScore W2797798064C41008148 @default.
• W2797798064 hasConceptScore W2797798064C50644808 @default.
• W2797798064 hasConceptScore W2797798064C81363708 @default.
• W2797798064 hasLocation W27977980641 @default.
• W2797798064 hasOpenAccess W2797798064 @default.
• W2797798064 hasPrimaryLocation W27977980641 @default.
• W2797798064 hasRelatedWork W164697790 @default.
• W2797798064 hasRelatedWork W2024736329 @default.
• W2797798064 hasRelatedWork W2096011958 @default.
• W2797798064 hasRelatedWork W2118527271 @default.
• W2797798064 hasRelatedWork W2558621277 @default.
• W2797798064 hasRelatedWork W2606554264 @default.
• W2797798064 hasRelatedWork W2617670119 @default.
• W2797798064 hasRelatedWork W2766678531 @default.
• W2797798064 hasRelatedWork W278373674 @default.
• W2797798064 hasRelatedWork W2790283733 @default.
• W2797798064 hasRelatedWork W2905600194 @default.
• W2797798064 hasRelatedWork W2906361042 @default.
• W2797798064 hasRelatedWork W2914114513 @default.
• W2797798064 hasRelatedWork W2946501903 @default.
• W2797798064 hasRelatedWork W2962734330 @default.
• W2797798064 hasRelatedWork W3016928826 @default.
• W2797798064 hasRelatedWork W3031618159 @default.
• W2797798064 hasRelatedWork W3082942429 @default.
• W2797798064 hasRelatedWork W3160683869 @default.
• W2797798064 hasRelatedWork W69316309 @default.
• W2797798064 isParatext "false" @default.
• W2797798064 isRetracted "false" @default.
• W2797798064 magId "2797798064" @default.

• next