Design of Oscillator Networks for Generating Signal with Prescribed Statistical Property

Abstract

Abstract. We design oscillator networks to generate a signal with a prescribed power spectrum. We consider networks of identical sin-wave oscillators and the Kuramoto order parameter as an output signal of the system. We use the Kullback-Leibler entropy as a measure of the distance between the power spectrum of the output signal and those of desired one. By optimizing the connection network through the Markov chain Monte Carlo method with replica exchange, we found that even oscillator network with a small number of elements can be generate a variety of time signals. The output signals include periodic and quasi-periodic signals with prescribed periods, and aperiodic signals with prescribed power spectrums.