Matches in SemOpenAlex for { <https://semopenalex.org/work/W2896884791> ?p ?o ?g. }
- W2896884791 abstract "Product unit neural networks (PUNNs) are powerful representational models with a strong theoretical basis, but have proven to be difficult to train with gradient-based optimizers. We present windowed product unit neural networks (WPUNNs), a simple method of leveraging product as a nonlinearity in a neural network. Windowing the product tames the complex gradient surface and enables WPUNNs to learn effectively, solving the problems faced by PUNNs. WPUNNs use product layers between traditional sum layers, capturing the representational power of product units and using the product itself as a nonlinearity. We find the result that this method works as well as traditional nonlinearities like ReLU on the MNIST dataset. We demonstrate that WPUNNs can also generalize gated units in recurrent neural networks, yielding results comparable to LSTM networks." @default.
- W2896884791 created "2018-10-26" @default.
- W2896884791 creator A5001526908 @default.
- W2896884791 creator A5003595274 @default.
- W2896884791 date "2018-10-01" @default.
- W2896884791 modified "2023-09-24" @default.
- W2896884791 title "Leveraging Product as an Activation Function in Deep Networks" @default.
- W2896884791 cites W1511704128 @default.
- W2896884791 cites W1598265492 @default.
- W2896884791 cites W1810943226 @default.
- W2896884791 cites W1924770834 @default.
- W2896884791 cites W1925618211 @default.
- W2896884791 cites W1974719097 @default.
- W2896884791 cites W2005708641 @default.
- W2896884791 cites W2017880785 @default.
- W2896884791 cites W2018444051 @default.
- W2896884791 cites W2040370888 @default.
- W2896884791 cites W2064052975 @default.
- W2896884791 cites W2064675550 @default.
- W2896884791 cites W2070450685 @default.
- W2896884791 cites W2079044132 @default.
- W2896884791 cites W2080929898 @default.
- W2896884791 cites W2081070226 @default.
- W2896884791 cites W2085232632 @default.
- W2896884791 cites W2100380271 @default.
- W2896884791 cites W2108909252 @default.
- W2896884791 cites W2109527845 @default.
- W2896884791 cites W2116261113 @default.
- W2896884791 cites W2121029939 @default.
- W2896884791 cites W2126003254 @default.
- W2896884791 cites W2130937797 @default.
- W2896884791 cites W2157331557 @default.
- W2896884791 cites W2163605009 @default.
- W2896884791 cites W2170942820 @default.
- W2896884791 cites W2266238587 @default.
- W2896884791 cites W2290998037 @default.
- W2896884791 cites W2294797155 @default.
- W2896884791 cites W2327617216 @default.
- W2896884791 cites W2401697546 @default.
- W2896884791 cites W2498118637 @default.
- W2896884791 cites W2512749620 @default.
- W2896884791 doi "https://doi.org/10.1109/smc.2018.00280" @default.
- W2896884791 hasPublicationYear "2018" @default.
- W2896884791 type Work @default.
- W2896884791 sameAs 2896884791 @default.
- W2896884791 citedByCount "0" @default.
- W2896884791 crossrefType "proceedings-article" @default.
- W2896884791 hasAuthorship W2896884791A5001526908 @default.
- W2896884791 hasAuthorship W2896884791A5003595274 @default.
- W2896884791 hasBestOaLocation W28968847912 @default.
- W2896884791 hasConcept C108583219 @default.
- W2896884791 hasConcept C111472728 @default.
- W2896884791 hasConcept C121332964 @default.
- W2896884791 hasConcept C138885662 @default.
- W2896884791 hasConcept C14036430 @default.
- W2896884791 hasConcept C154945302 @default.
- W2896884791 hasConcept C158622935 @default.
- W2896884791 hasConcept C190502265 @default.
- W2896884791 hasConcept C2524010 @default.
- W2896884791 hasConcept C2780586882 @default.
- W2896884791 hasConcept C2984842247 @default.
- W2896884791 hasConcept C33923547 @default.
- W2896884791 hasConcept C38365724 @default.
- W2896884791 hasConcept C41008148 @default.
- W2896884791 hasConcept C50644808 @default.
- W2896884791 hasConcept C62520636 @default.
- W2896884791 hasConcept C78458016 @default.
- W2896884791 hasConcept C86803240 @default.
- W2896884791 hasConcept C90673727 @default.
- W2896884791 hasConceptScore W2896884791C108583219 @default.
- W2896884791 hasConceptScore W2896884791C111472728 @default.
- W2896884791 hasConceptScore W2896884791C121332964 @default.
- W2896884791 hasConceptScore W2896884791C138885662 @default.
- W2896884791 hasConceptScore W2896884791C14036430 @default.
- W2896884791 hasConceptScore W2896884791C154945302 @default.
- W2896884791 hasConceptScore W2896884791C158622935 @default.
- W2896884791 hasConceptScore W2896884791C190502265 @default.
- W2896884791 hasConceptScore W2896884791C2524010 @default.
- W2896884791 hasConceptScore W2896884791C2780586882 @default.
- W2896884791 hasConceptScore W2896884791C2984842247 @default.
- W2896884791 hasConceptScore W2896884791C33923547 @default.
- W2896884791 hasConceptScore W2896884791C38365724 @default.
- W2896884791 hasConceptScore W2896884791C41008148 @default.
- W2896884791 hasConceptScore W2896884791C50644808 @default.
- W2896884791 hasConceptScore W2896884791C62520636 @default.
- W2896884791 hasConceptScore W2896884791C78458016 @default.
- W2896884791 hasConceptScore W2896884791C86803240 @default.
- W2896884791 hasConceptScore W2896884791C90673727 @default.
- W2896884791 hasLocation W28968847911 @default.
- W2896884791 hasLocation W28968847912 @default.
- W2896884791 hasOpenAccess W2896884791 @default.
- W2896884791 hasPrimaryLocation W28968847911 @default.
- W2896884791 hasRelatedWork W2201721810 @default.
- W2896884791 hasRelatedWork W2604802223 @default.
- W2896884791 hasRelatedWork W2767072113 @default.
- W2896884791 hasRelatedWork W2860351735 @default.
- W2896884791 hasRelatedWork W2861440650 @default.
- W2896884791 hasRelatedWork W2896884791 @default.
- W2896884791 hasRelatedWork W2947175736 @default.
- W2896884791 hasRelatedWork W2949560545 @default.