A magnonic-optoelectronic reservoir for physical reservoir computing

Abstract

Physical reservoir computing is a promising approach for fast and energy efficient computer vision, natural language processing, and general pattern recognition. This work presents a physical reservoir based on magnonic-optoelectronic oscillator (MOEO). This approach allowed us to realize short-term memory and nonlinearity as separate system components. The device`s optical path uses a fiber-optic delay line as a short-term-memory element. The microwave path is responsible for nonlinear mapping of input data to a higher-dimensional space. The strong four-wave nonlinearity of spin waves propagating in an yttrium-iron garnet (YIG) ferrite film enables the process. The reservoir performance is evaluated by completing task-independent tests known as short-term memory (STM) and parity-check (PC) tasks. In addition, a numerical model of the MOEO based reservoir is developed. Results of the numerical simulation of the reservoir performance are in good agreement with the experimental data.

References in corpus (6)

• Recent Advances in Physical Reservoir Computing: A Review
• Optoelectronic Reservoir Computing
• Reservoir computing using a spin-wave delay line active ring resonator
• Reservoir computing based on a silicon microring and time multiplexing for binary and analog operations
• Implementing a magnonic time-delay reservoir computer model
• Experimental parameters, combined dynamics, and nonlinearity of a Magnonic-Opto-Electronic Oscillator (MOEO)