CONTINUOUS HOPFIELD COMPUTATIONAL NETWORK - HARDWARE IMPLEMENTATION

A simple continuous type of Hopfield network is studied and the principle behind the operation of these circuits is discussed. These circuits are also known as collective decision circuits. The nature or such circuits suggests that they are more effective for problems that involve global interaction between different parts of the problem. The neural network that can solve the problem has many interconnected amplifiers. It is able rapidly to follow the computation path to a solution. In order to examine the basic principles of all such complicated circuits, we consider the design fabrication and performance evaluation of a simple flip-flop circuit popularly used in the electronics industry. The circuit uses general principles resulting from the understanding of the basic collective computational properties of a specific class of analogue Hopfield network. The circuit is tested for different input parameters and computational energy computed in each case. The process is repeated with external input current (bias). The basic idea is to demonstrate the computational power and the speed of collective analogue networks of neurons in solving the optimization problem rapidly. © 1990 Taylor & Francis Ltd.