Randomly generated node-state-update procedure for dipole-coupled magnetic reservoir computing with voltage control of the magnetism

We demonstrate reservoir computing using a dipole-coupled magnetic-dot array by simulations with a macro-spin model at 0 K. The magnetic moment of the dot is used for the node state of the reservoir. We perform short-term-memory (STM) and parity check (PC) tasks to evaluate the performance of the reservoir. Binary information was written to the input magnetic dots as a direction of the magnetic moment. A voltage- induced magnetic-anisotropy change was used to update the node state. We investigated the STM capacity (STMC) and PC capacity (PCC)dependence on the number of magnetic dots by means of randomly generated node-update procedures. As a result, the STMC and PCC are strongly dependent on the nodestate update procedure.