Memristive discrete chaotic neural network and its application in associative memory

Abstract

Chaotic behaviors existing in biological neurons play an important role in the brain’s associative memory. Hence, chaotic neural networks have been widely applied in associative memory. This paper proposed a discrete chaotic neural network which is implemented by electronic components not by computer software. This chaotic neural network is a Hopfield neural network consisting of synapses and chaotic neurons. The realization of synapses is based on a memristive crossbar array and operational amplifiers. By adjusting the value of memristance, the synaptic weights with positive, negative, and zero values are realized. The chaotic neuron is composed of operational amplifiers and voltage-controlled switches, and it can generate chaotic signals and finish the iterative operation of the system. A chaotic neural network with 9 neurons is constructed as an example, and the influence of different initial states on the multi-associative memory is investigated. The simulation results demonstrate the single-associative and multi-associative memories of the proposed chaotic neural network.