Simulation Of Logic Elements In Reverse Mode For Building Neural Networks

The hardware implementation of a feedback neural network, which is based on two ideas: Stephen Grasberg's adaptive resonance theory and Hopfield's auto-associative memory, requires that all elements be connected by direct and feedback connections. This paper presents a simulation of ordinary logic elements, but with reversible properties. The proposed structure allows the signals to move both in the forward and the opposite directions. In practice, this will mean that with the help of control signals, neuron signals can be directed along the same lines of communication in different directions. In this way, all elements of the system will be interconnected and the system will be able to remember and choose the best way to solve the problem.The authors propose modeling of ordinary logic elements using a special inclusion scheme based on resistive dividers, which allows providing their reversible mode of operation. It is also proposed to use the following logical elements to build neural networks such as the Cosco architecture.The work simulates the reversible operation of the following logical elements: 1) repeater (buffer); 2) inverter; 3) the logical element NAND; 4) logical element NOR.