1.
Med Tekh
; (2): 15-8, 2008.
Artigo
em Russo
| MEDLINE
| ID: mdl-18507134
RESUMO
Problems of MATLAB-based simulation of hybrid neural networks for data classification in biomedical applications are considered. The neural-network structure is used as a defuzzifier. Graphic interfaces for constructing fuzzy neural-network models are suggested. Methods for merging models constructed using different MATLAB packages and an algorithm for constructing a fuzzy neural-network model are suggested. The algorithm includes both modules for tuning membership functions of the fuzzy system fuzzifier, and modules for tuning the parameters of the defuzzifier implemented as a two-layer neural-network structure of the perceptron type.