Memristive Hopfield Neural Network for Reasoning with Incomplete Information and Its Circuit Implementation

Abstract

Memristor-based neural networks have been extensively studied, but reasoning as an important topic of artificial intelligence is rarely implemented directly by circuit. Reasoning, as an important part of artificial intelligence, is an open and challenging problem to be solved. In this paper, memristive hopfield neural network is designed to realize reasoning. The designed circuit consists of four modules, namely a signal processing module, an iterator module, a signal input module and a signal output module. The signal processing module performs iterative operations under the control of the iterator module, so that the output signals of memristive hopfield neural network can converge to the final states. Reasoning is one of the basic forms of thinking, and is the process of drawing result from one or several given conditions. A guessing game for athletes is completed by the designed circuit which can reason the name of the athlete from incomplete information. The simulation results verify the feasibility of the circuit for reasoning.