Memristor-based neural network circuit with weighted sum simultaneous perturbation training and its applications

In this work, a full circuit of memristor-based neural network with weighted sum simultaneous perturbation training is proposed. Firstly, a synaptic circuit is designed by using a pair of memristors, which can represent negative, zero, and positive synaptic weights. Secondly, a full circuit of the neural network is designed, with all operations being completed on the circuit without any computer aid. The neural network is trained with the weighted sum simultaneous perturbation algorithm. The algorithm does not involve complex derivative calculation and error back propagation, and it only applies perturbations to weighted sum, so the circuit implementation is more simple. Finally, application simulations of the proposed neural network circuit are performed via PSpice. The results of simulation indicate that the memristor-based neural network is practical and effective. (c) 2021 Elsevier B.V. All rights reserved.

Automated summary

In this work, a memristor-based neural network circuit with weighted sum simultaneous perturbation training is proposed, which simplifies the training process and achieves practical and effective results. The circuit design is efficient and eliminates the need for complex computations, showing promising potential for neural network applications.