Accelerated Easy Learning Method for Nonlinear Regression

摘要

The present invention relates to a learning method for identifying the dynamics of data by using an estimation network consisting of three layers, that is, an input layer, a hidden layer, and an output layer, according to a given approximation degree for continuously input data.;Here, the input and output layers of the estimation network are composed of linear functions, whereas the hidden layer is composed of Gaussian element functions.;The proposed accelerated learning method has the advantages of comparable computational time and good parameter estimation.;As a result of applying the proposed learning method to the M-G chaotic time series predictor, we obtained better predictions than the M-G chaotic time series predictor using the simple learning method.