Global Dynamics of a Stochastic Neuronal Oscillator

TL;DR

This paper analyzes how the global dynamics of a stochastic neuronal oscillator depend on relaxation rate, noise, and input parameters using Markov operators, revealing insights into spike timing and past activity influence.

Contribution

It introduces a Markov operator framework for the infinite relaxation rate case, extending previous models to better understand stochastic neuronal oscillators' responses.

Findings

The response to impulses can be described by a product of Markov operators.

Eigenvalue analysis decomposes responses into stationary and transient components.

Past activity duration depends on relaxation rate, noise, and input parameters.

Abstract

Nonlinear oscillators have been used to model neurons that fire periodically in the absence of input. These oscillators, which are called neuronal oscillator, share some common response structures with other biological oscillations. In this study, we analyze the dependence of the global dynamics of an impulse-driven stochastic neuronal oscillator on the relaxation rate to the limit cycle, the strength of the intrinsic noise, and the impulsive input parameters. To do this, we use a Markov operator that both reflects the density evolution of the oscillator and is an extension of the phase transition curve, which describes the phase shift due to a single isolated impulse. Previously, we derived the Markov operator for the finite relaxation rate that describes the dynamics of the entire phase plane. Here, we construct a Markov operator for the infinite relaxation rate that describes the stochastic dynamics restricted to the limit cycle. In both cases, the response of the stochastic neuronal oscillator to time-varying impulses is described by a product of Markov operators. Furthermore, we calculate the number of spikes between two consecutive impulses to relate the dynamics of the oscillator to the number of spikes per unit time and the interspike interval density. Specifically, we analyze the dynamics of the number of spikes per unit time based on the properties of the Markov operators. Each Markov operator can be decomposed into stationary and transient components based on the properties of the eigenvalues and eigenfunctions. This allows us to evaluate the difference in the number of spikes per unit time between the stationary and transient responses of the oscillator, which we show to be based on the dependence of the oscillator on past activity. Our analysis shows how the duration of the past neuronal activity depends on the relaxation rate, the noise strength and the input parameters.