Macromagnetic Simulation for Reservoir Computing Utilizing Spin Dynamics in Magnetic Tunnel Junctions

Abstract

The recurrent neural network, a machine-learning approach, is a mathematical model that emulates neuronal function in the human brain. The authors report a quantitative analysis of the figures of merit for reservoir computing, which is a type of recurrent neural network, using the spintronic devices known as magnetic tunnel junctions (MTJs). While MTJs are usually investigated in the context of high-density nonvolatile digital storage, these results show that they are also suitable for advanced computation.