誤差の変化率を考慮した階層ニューラルネットワークの学習アルゴリズム

概要

論文の詳細を見る
In this paper, we propose a fast learning algorithm for the multi-layered neural network which converges the sum total of the error quickly. The algorithm considers especially the variation rate of repetition nth and (n-1)th error between the output signal on each unit of output layer and the teacher signal (to be compared with the output signal). The proposed algorithm is an extention of conventional back-propagation learning algorithm. This learning algorithm makes possible the acceleration of learning, one of weak points on back-propagation, and makes more acceleration possible of conventional back-propagation by the introduction of the proposed learning algorithm into the various reports concerning acceleration (of conventional back-propagation). The effectiveness of this method on convergence of the sum total of the error has been verified by the computer simulation for two examples of the exclusive-OR and the cosine function using proposed learning algorithm.
東海大学の論文
1995-03-30