Physical Reservoir Computing Using Nonlinear MEMS Resonator Having High Memory Capacity at "Edge of Chaos"

Abstract

This paper reports physical reservoir computing (PRC) using a single nonlinear electrostatic resonator and demonstrates its high memory capacity at "edge of chaos." The resonator is a simple doubly supported resonator fabricated from a silicon-on-insulator wafer. We proposed a PRC system without feedback loop, in which its memory capacity relies on the decay time of the high-Q resonator. The benchmark task results indicate that the system shows good linear and nonlinear memory capacities at the resonance and the maximum capacity was obtained at the vicinity of the instability edge of the frequency response.