Learning Method of Nonlinear Estimation Network for Approximation of Nonlinear Function

摘要

The present invention relates to a method for approximating a nonlinear function using a three-layer structure, that is, an estimation network composed of an input layer, a hidden layer and an output layer.;Here, the input layer and the output layer are made up of a linear function, while the hidden layer consists of an element function having limitation and localization.;Using this estimation network, the nonlinear function implied by the given learning data is approximated.;Here, the estimation process, that is, the learning process is divided into two parts. In the first part, the number of appropriate element functions is determined according to the degree of function approximation given. In the second part, the parameters of a given network are estimated. In the estimation method, Applies a linear learning method, and applies a nonlinear (partial linear) learning method between the hidden layer and the input layer.;As a result, the proposed learning process is superior to other similar nonlinear function approximation methods in terms of number of element functions and learning speed.