FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2056517685> ?p ?o ?g. }

• W2056517685 endingPage "7609" @default.
• W2056517685 startingPage "7590" @default.
• W2056517685 abstract "A two-layer feedforward network of McCulloch-Pitts neurons with N inputs and K hidden units is analyzed for Nensuremath{rightarrow}ensuremath{infty} and K finite with respect to its ability to implement p=ensuremath{alpha}N random input-output relations. Special emphasis is put on the case where all hidden units are coupled to the output with the same strength (committee machine) and the receptive fields of the hidden units either enclose all input units (fully connected) or are nonoverlapping (tree structure). The storage capacity is determined generalizing Gardner's treatment [J. Phys. A 21, 257 (1988); Europhys. Lett. 4, 481 (1987)] of the single-layer perceptron. For the treelike architecture, a replica-symmetric calculation yields ${mathrm{ensuremath{alpha}}}_{mathit{c}}$ensuremath{propto} ensuremath{surd}K for a large number K of hidden units. This result violates an upper bound derived by Mitchison and Durbin [Biol. Cybern. 60, 345 (1989)]. One-step replica-symmetry breaking gives lower values of ${mathrm{ensuremath{alpha}}}_{mathit{c}}$.In the fully connected committee machine there are in general correlations among different hidden units. As the limit of capacity is approached, the hidden units are anticorrelated: One hidden unit attempts to learn those patterns which have not been learned by the others. These correlations decrease as 1/K, so that for Kensuremath{rightarrow}ensuremath{infty} the capacity per synapse is the same as for the tree architecture, whereas for small K we find a considerable enhancement for the storage per synapse. Numerical simulations were performed to explicitly construct solutions for the tree as well as the fully connected architecture. A learning algorithm is suggested. It is based on the least-action algorithm, which is modified to take advantage of the two-layer structure. The numerical simulations yield capacities p that are slightly more than twice the number of degrees of freedom, while the fully connected net can store relatively more patterns than the tree. Various generalizations are discussed. Variable weights from hidden to output give the same results for the storage capacity as does the committee machine, as long as K=O(1). We furthermore show that thresholds at the hidden units or the output unit cannot increase the capacity, as long as random unbiased patterns are considered. Finally we indicate how to generalize our results to other Boolean functions." @default.
• W2056517685 created "2016-06-24" @default.
• W2056517685 creator A5004034903 @default.
• W2056517685 creator A5014558541 @default.
• W2056517685 creator A5028489124 @default.
• W2056517685 creator A5053902486 @default.
• W2056517685 creator A5072436931 @default.
• W2056517685 date "1992-05-01" @default.
• W2056517685 modified "2023-10-14" @default.
• W2056517685 title "Storage capacity and learning algorithms for two-layer neural networks" @default.
• W2056517685 cites W1498436455 @default.
• W2056517685 cites W1972884424 @default.
• W2056517685 cites W1973670813 @default.
• W2056517685 cites W1974572752 @default.
• W2056517685 cites W1979005517 @default.
• W2056517685 cites W1995539593 @default.
• W2056517685 cites W1997011744 @default.
• W2056517685 cites W2001570872 @default.
• W2056517685 cites W2009784977 @default.
• W2056517685 cites W2010833108 @default.
• W2056517685 cites W2012903341 @default.
• W2056517685 cites W2013128153 @default.
• W2056517685 cites W2028923186 @default.
• W2056517685 cites W2030450972 @default.
• W2056517685 cites W2033606703 @default.
• W2056517685 cites W2043014754 @default.
• W2056517685 cites W2044767119 @default.
• W2056517685 cites W2055417373 @default.
• W2056517685 cites W2077101514 @default.
• W2056517685 cites W2080531309 @default.
• W2056517685 cites W2088522405 @default.
• W2056517685 cites W2161278885 @default.
• W2056517685 cites W2788237242 @default.
• W2056517685 doi "https://doi.org/10.1103/physreva.45.7590" @default.
• W2056517685 hasPubMedId "https://pubmed.ncbi.nlm.nih.gov/9906832" @default.
• W2056517685 hasPublicationYear "1992" @default.
• W2056517685 type Work @default.
• W2056517685 sameAs 2056517685 @default.
• W2056517685 citedByCount "62" @default.
• W2056517685 countsByYear W20565176852012 @default.
• W2056517685 countsByYear W20565176852013 @default.
• W2056517685 countsByYear W20565176852014 @default.
• W2056517685 countsByYear W20565176852015 @default.
• W2056517685 countsByYear W20565176852019 @default.
• W2056517685 countsByYear W20565176852021 @default.
• W2056517685 countsByYear W20565176852022 @default.
• W2056517685 countsByYear W20565176852023 @default.
• W2056517685 crossrefType "journal-article" @default.
• W2056517685 hasAuthorship W2056517685A5004034903 @default.
• W2056517685 hasAuthorship W2056517685A5014558541 @default.
• W2056517685 hasAuthorship W2056517685A5028489124 @default.
• W2056517685 hasAuthorship W2056517685A5053902486 @default.
• W2056517685 hasAuthorship W2056517685A5072436931 @default.
• W2056517685 hasConcept C113174947 @default.
• W2056517685 hasConcept C11413529 @default.
• W2056517685 hasConcept C114614502 @default.
• W2056517685 hasConcept C121332964 @default.
• W2056517685 hasConcept C122637931 @default.
• W2056517685 hasConcept C134306372 @default.
• W2056517685 hasConcept C145420912 @default.
• W2056517685 hasConcept C154945302 @default.
• W2056517685 hasConcept C33923547 @default.
• W2056517685 hasConcept C41008148 @default.
• W2056517685 hasConcept C50644808 @default.
• W2056517685 hasConcept C60908668 @default.
• W2056517685 hasConcept C77553402 @default.
• W2056517685 hasConceptScore W2056517685C113174947 @default.
• W2056517685 hasConceptScore W2056517685C11413529 @default.
• W2056517685 hasConceptScore W2056517685C114614502 @default.
• W2056517685 hasConceptScore W2056517685C121332964 @default.
• W2056517685 hasConceptScore W2056517685C122637931 @default.
• W2056517685 hasConceptScore W2056517685C134306372 @default.
• W2056517685 hasConceptScore W2056517685C145420912 @default.
• W2056517685 hasConceptScore W2056517685C154945302 @default.
• W2056517685 hasConceptScore W2056517685C33923547 @default.
• W2056517685 hasConceptScore W2056517685C41008148 @default.
• W2056517685 hasConceptScore W2056517685C50644808 @default.
• W2056517685 hasConceptScore W2056517685C60908668 @default.
• W2056517685 hasConceptScore W2056517685C77553402 @default.
• W2056517685 hasIssue "10" @default.
• W2056517685 hasLocation W20565176851 @default.
• W2056517685 hasLocation W20565176852 @default.
• W2056517685 hasOpenAccess W2056517685 @default.
• W2056517685 hasPrimaryLocation W20565176851 @default.
• W2056517685 hasRelatedWork W1501774291 @default.
• W2056517685 hasRelatedWork W2031152868 @default.
• W2056517685 hasRelatedWork W2073377738 @default.
• W2056517685 hasRelatedWork W2085270790 @default.
• W2056517685 hasRelatedWork W2161649813 @default.
• W2056517685 hasRelatedWork W2392110728 @default.
• W2056517685 hasRelatedWork W2592261830 @default.
• W2056517685 hasRelatedWork W2809374263 @default.
• W2056517685 hasRelatedWork W2982347821 @default.
• W2056517685 hasRelatedWork W2992225777 @default.
• W2056517685 hasVolume "45" @default.
• W2056517685 isParatext "false" @default.
• W2056517685 isRetracted "false" @default.
• W2056517685 magId "2056517685" @default.

• next