Binary Factorization in Hopfield-Like Neural Autoassociator: A Promising Tool for Data Compression

0404887 - UIVT-O 20030184 RIV AT eng C - Conference Paper (international conference)
Frolov, A. A. - Sirota, A.M. - Húsek, Dušan - Muraviev, I. - Combe, P.
Binary Factorization in Hopfield-Like Neural Autoassociator: A Promising Tool for Data Compression.
Artificial Neural Nets and Genetic Algorithms. Wien: SpringerVerlag, 2003 - (Pearson, D.; Steele, N.; Albrecht, R.), s. 58-62. ISBN 3-211-00743-1.
[ICANNGA'2003 /6./. Roanne (FR), 23.04.2003-25.04.2003]
R&D Projects: GA ČR GA201/01/1192
Grant - others:FR-CZ Barrande(XX) 04693QG
Institutional research plan: AV0Z1030915
Keywords : neural networks * binary factorization * data compression
Subject RIV: JD - Computer Applications, Robotics

Proposed approach of data compression is based on feature extraction procedure which maps original patterns into features (factors) space of reduced, possibily very small, dimension. It is shown that Hebbian unsupervised learning of Hopfield-like neural network is a natural procedure for factor extraction. Due to this learning, factors become the attractors of network dynamics, hence they can be revealed by the random search.
Permanent Link: http://hdl.handle.net/11104/0125111