WEIGHTED MAJORITY DECISION PROCESSING SYSTEM BY NEURAL NETWORK

摘要

PURPOSE:To output an estimation value with less errors by transmitting the output of a neural network and the output of a weight value addition means to a division means, obtaining the quotient and outputting the estimation value. CONSTITUTION:An input signal is mask-controlled lest it adds '0' of a weight value by using the neural network 1 which is learnt and adjusted, and the sum SIGMAWi of the weight values except for the weight value is outputted from the weight addition part 5. Then, XSIGMAWj and SIGMAWj outputted from the neural network 1 are divided in the division means 6 and the estimation value X is obtained. Thus, an operation can be executed in a majority decision graph surface without a discontinuous point. Thus, the value which the neural network can easily learn and which has less errors can be obtained.