Presentation 2003/6/20
On the pattern representation self-organized by a model based on a randomly and symmetrically connected network
Yasuhiro HIRATA, Kouji KURATA, Naoki OSHIRO,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) A new self-organizing model is studied numerically. Preceding self-organizing models of topological mappings, such as von der Malsburg's and Amari's, are based on a neural array with mexican-hat-type connection pattern. In our model, a representation layer consisting of neural units with random and symmetric, fixed connections is used instead of the neural array. The connections from an input layer to the representation layer are modified by Hebbian rule with regularization. The model assigns to an input a representation that is a firing pattern on the representation layer. It is shown that the representation preserves the distance relation between the inputs to some extent, and has finer resolution for frequent inputs. Detail relationship between input frequency and weight vector density after learning is numerically studied.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) self-organizing model / random by connected work / information representation
Paper # NC2003-18
Date of Issue

Conference Information
Committee NC
Conference Date 2003/6/20(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Neurocomputing (NC)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) On the pattern representation self-organized by a model based on a randomly and symmetrically connected network
Sub Title (in English)
Keyword(1) self-organizing model
Keyword(2) random by connected work
Keyword(3) information representation
1st Author's Name Yasuhiro HIRATA
1st Author's Affiliation Faculty of Engineering, University of the Ryukyus()
2nd Author's Name Kouji KURATA
2nd Author's Affiliation Faculty of Engineering, University of the Ryukyus
3rd Author's Name Naoki OSHIRO
3rd Author's Affiliation Faculty of Engineering, University of the Ryukyus
Date 2003/6/20
Paper # NC2003-18
Volume (vol) vol.103
Number (no) 153
Page pp.pp.-
#Pages 6
Date of Issue