拡張カルマンフィルタによる超高速ニューロ学習 : 第1報, 排他的論理和問題への適用

抄録

The development of a backpropagation (BP) algorithm for neuron training has made it possible to use the layered neural network for simulating the nonlinear system. On the other hand, the neuron training algorithm based on the extended Kalman filter algorithm (KNT) has been developed. It is already shown that KNT has high performance compared with the BP. In this report, we show a technique to accelerate the training speed of the KNT. We call this technique the VC (variable covariance) technique. By applying the KNT with the VC technique to an exclusive OR problem, we clarify the performance of the present acceleration technique.