Universal scaling between wave speed and size enables nanoscale high-performance reservoir computing based on propagating spin-waves
Neuromorphic computing using spin waves is promising for high-speed nanoscale devices, but the realization of high performance has not yet been achieved. Here we show, using micromagnetic simulations and simplified theory with response functions, that spin-wave physical reservoir computing can achieve miniaturization down to nanoscales keeping high computational power comparable with other state-of-art systems. We also show the scaling of system sizes with the propagation speed of spin waves plays a key role to achieve high performance at nanoscales.