Summary
International Symposium on Nonlinear Theory and its Applications
2008
Session Number:B4L-D
Session:
Number:B4L-D4
Data Mining CNN to Circuit Modeling
Mamoru Tanaka, Yuko Zennyoji, Hisashi Aomori,
pp.-
Publication Date:2008/9/7
Online ISSN:2188-5079
Summary:
This paper describes a machine learning of a data-mining cellular neural network (CNN) from analysis of covariance structure using Back Euler method. It is the implicit method which is the most practical method for solving stiff systems. At each learning iteration step for quasi-Newton method, the Hessian matrix approximation for the matrix including second-order partial derivatives is updated by using Davidon-Fletcher-Powell (DFP) method. By using both of Back Euler method and DFP method, the solution for stiff systems in the parameter space can be obtained. That is, our purpose is to determine the weight parameters θ in the connection matrices A, B, C, D, T and e by the machine learning method for equilibrium points of the CNN states equation x ? = 0. The structure of the data-mining CNN is considered from viewpoints of circuit modeling.