Holographic implementation of a Hopfield model with discrete weightings

This paper investigates the feasibility of constructing a Hopfield neural network using optical techniques. The particular implementation utilizes parallel holographic interconnections and could have a processing speed vastly greater than the serial electronic computers we use today. This improvement in speed is of great importance, particularly in the field of computer vision. It is found that present day technology would seriously limit the size of the network that could be implemented. Our analysis suggests, however, that the construction of an optical machine is now possible with a connectivity exceeding any electronic machine available.