Memristor modeling with homeostatic threshold variation for simulation and application

This paper proposes a novel approach to emulating biological neural components using memristors. The nanoscale, resistance variability, and non-volatility of the memristor make it a suitable candidate for this task. A homeostasis-type switching mechanism is induced into the memristor modeling, inspired by the homeostatic switching behavior of biological neural components. The proposed model is generalizable and adaptable to various threshold-type memristor models. A SPICE model is proposed, and the memristor model is utilized to implement a neuron circuit that follows the homeostasis-type mechanism. A spiking neural network and its verification platform are designed, where pattern recognition can be achieved. The simulation is carried out in Cadence PSPICE, and a test comparison between scenarios with homeostasis and non-homeostasis is proposed. This work demonstrates the potential of memristors in implementing neuromorphic computing systems.