液晶空間光変調素子を用いた光ニューラルコンピューティング

抄録

We present a new optoelectronic feed-forward neural network system which is able to train the network according to delta-rule without using any external computers. This neural network system is composed of Spatial Light Modulators using Ferroelectric liquid Crystal. This optical system calculates the change of interconnection weights and automatically corrects the interconnection weights. The basic operations of this neural network system, learning and pattern classification, are successfully demonstrated.