Real neuronal networks in the brain utilize networks of neurons with graded not binary firing rates. A theoretical analysis of the operation of autoassociative networks with neurons with graded firing rates has therefore been developed. The present simulation study was performed in order to investigate the operation of such a network with values for the asymmetric diluted neuronal connectivity typical of some brain regions such as the hippocampus, which are outside the range to which the theoretical analysis strictly applies. We report that, in line with theoretical predictions, the amount of information that can be retrieved is relatively independent of the resolution of the stored patterns (binary, ternary, decimal, or fifty-fold). The implication of this is that if the network stores many graded patterns, which it can, then the retrieval quality of each of he patterns becomes low. The implications of this trade-off between the number of patterns stored and the retrieval quality of each pattern when graded firing rates are stored for understanding he operation of networks in the hippocampus are considered.

Simulation studies of the CA3 hippocampal subfield modelled as an attractor neural network / Rolls, Edmund T.; Treves, Alessandro; Foster, David; Perez-Vicente, Conrado. - In: NEURAL NETWORKS. - ISSN 0893-6080. - 10:9(1997), pp. 1559-1569. [10.1016/S0893-6080(97)00092-0]

Simulation studies of the CA3 hippocampal subfield modelled as an attractor neural network

Rolls, Edmund T.;Treves, Alessandro;
1997

Abstract

Real neuronal networks in the brain utilize networks of neurons with graded not binary firing rates. A theoretical analysis of the operation of autoassociative networks with neurons with graded firing rates has therefore been developed. The present simulation study was performed in order to investigate the operation of such a network with values for the asymmetric diluted neuronal connectivity typical of some brain regions such as the hippocampus, which are outside the range to which the theoretical analysis strictly applies. We report that, in line with theoretical predictions, the amount of information that can be retrieved is relatively independent of the resolution of the stored patterns (binary, ternary, decimal, or fifty-fold). The implication of this is that if the network stores many graded patterns, which it can, then the retrieval quality of each of he patterns becomes low. The implications of this trade-off between the number of patterns stored and the retrieval quality of each pattern when graded firing rates are stored for understanding he operation of networks in the hippocampus are considered.
10
9
1559
1569
Rolls, Edmund T.; Treves, Alessandro; Foster, David; Perez-Vicente, Conrado
File in questo prodotto:
File Dimensione Formato  
Rol+97b.pdf

non disponibili

Tipologia: Versione Editoriale (PDF)
Licenza: Non specificato
Dimensione 783.12 kB
Formato Adobe PDF
783.12 kB Adobe PDF   Visualizza/Apri   Richiedi una copia

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/20.500.11767/85780
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 62
  • ???jsp.display-item.citation.isi??? 53
social impact