Features Extraction by Hopfield-Like Neural Network

0405156 - UIVT-O 20030193 RIV ES eng C - Conference Paper (international conference)
Frolov, A. A. - Húsek, Dušan - Muraviev, I. P. - Řezanková, H. - Snášel, Václav - Polyakov, P.Y.
Features Extraction by Hopfield-Like Neural Network.
Neural Network Engineering Experiences. Malaga: Dpt. de Ingeneria de Sistemas y Automatica, 2003 - (Fernandez de Canete, J.; Tsaptsinos, D.), s. 383-390. ISBN 84-930984-1-8.
[EANN'03. International Conference on Engineering Applications of Neural Networks. Malaga (ES), 08.09.2003-10.09.2003]
R&D Projects: GA AV ČR IAA2030801; GA ČR GA201/01/1192
Grant - others:BARRANDE(XE) 203-030-2
Institutional research plan: AV0Z1030915
Keywords : Boolean factorization * recurrent neural network * single-step approximation
Subject RIV: BB - Applied Statistics, Operational Research

The unsupervised learning of feature extraction in high-dimensional patterns is a central problem for neural network approach. In this paper we demonstrate that Hebbian learning in Hopfield-like neural network is a natural procedure for feature extraction. Due to this learning, factors become attractors of network dynamics. The neurodynamics is analysed by Single-Step approximation and computer simulation.
Permanent Link: http://hdl.handle.net/11104/0125355