The application of the IIR-type neural network to the magnetic pole shape optimization

Abstract

A FIR type neural network that replaced the weight of the neural network with the FIR (Finite Impulse Response) type synapse was proposed previously, which is reported to be excellent for identification and prediction comparing with the usual FF (Feed Forward) type neural network. With reference to this idea, this paper presents the IIR (Infinite Impulse Response) type neural network that has delay elements to both input layer and output layer. We calculated the magnetic pole optimization using the proposed neural network. The result showed that the learning iteration could be reduced in comparison with the neural network of the usual FF.