Quantifying the Computational Capability of a Nanomagnetic Reservoir Computing Platform with Emergent Magnetization Dynamics

Devices based on arrays of interconnected magnetic nano-rings with emergent\nmagnetization dynamics have recently been proposed for use in reservoir\ncomputing applications, but for them to be computationally useful it must be\npossible to optimise their dynamical responses. Here, we use a phenomenological\nmodel to demonstrate that such reservoirs can be optimised for classification\ntasks by tuning hyperparameters that control the scaling and input rate of data\ninto the system using rotating magnetic fields. We use task-independent metrics\nto assess the rings' computational capabilities at each set of these\nhyperparameters and show how these metrics correlate directly to performance in\nspoken and written digit recognition tasks. We then show that these metrics,\nand performance in tasks, can be further improved by expanding the reservoir's\noutput to include multiple, concurrent measures of the ring arrays magnetic\nstates.\n