Character recognition using a dynamic optoelectronic neural network with unipolar binary weights

Abstract

A novel type of quantized learning rule with unipolar binary weights that is useful for the optical implementation of neural networks is reported. An input-dependent thresholding operation is also proposed to remove the unwanted effects that are due to the insufficient contrast ratio of spatial light modulators as synaptic connection devices. Moreover, we experimentally demonstrate the recognition of 26 characters of the alphabet by using the single set of an optoelectronic three-layered network.