Memristor-based Neural Network Circuit Simulation for Classification Application

With the advent of the big data era, the traditional von Neumann architecture, which has separate processing units and storage units, brings huge delay and energy consumption, and suffers from high data transmission costs. The memristor, a new principle device, has the ability to modulate the resistance based on the electrical flow passing through it, and its excellent nonlinear characteristics make it a basic component unit of a brain-inspired computing architecture. Scaling up the memristor and constructing a brain-inspired neural network architecture based on it, simulating the computing and learning process of artificial neural networks, is the fundamental solution to the von Neumann architecture bottleneck problem. In this paper, we use the LTspice circuit simulation software to build a brain-inspired neural network array based on memristor units and conduct target detection and classification application tests on typical data sets, thereby verifying the potential application of the memristive neural network architecture in the field of brain-inspired computing.