Oscillatory Survival Probability: Analytical, Numerical Study for oscillatory narrow escape and applications to neural network dynamics

TL;DR

This paper analyzes the oscillatory decay of survival probabilities in a stochastic escape problem near a focus, revealing a complex eigenvalue responsible for oscillations, with applications to neural network dynamics.

Contribution

It introduces the concept of oscillatory narrow escape, explicitly computes the second eigenvalue of the Fokker-Planck operator, and links oscillations to neural network behavior.

Findings

Complex eigenvalue causes oscillatory decay in survival probability.

Exit density concentrates on a small boundary region, indicating narrow escape.

Oscillations explain non-Poissonian escape times in neural models.

Abstract

We study the escape of Brownian motion from the domain of attraction $\Omega$ of a stable focus with a strong drift. The boundary $\partial \Omega$ of $\Omega$ is an unstable limit cycle of the drift and the focus is very close to the limit cycle. We find a new phenomenon of oscillatory decay of the peaks of the survival probability of the Brownian motion in $\Omega$. We compute explicitly the complex-valued second eigenvalue $\lambda_2(\Omega$) of the Fokker-Planck operator with Dirichlet boundary conditions and show that it is responsible for the peaks. Specifically, we demonstrate that the dominant oscillation frequency equals $1/{\mathfrak{I}}m\{\lambda_2(\Omega)\}$ and is independent of the relative noise strength. We apply the analysis to a canonical system and compare the density of exit points on $\partial \Omega$ to that obtained from stochastic simulations. We find that this density is concentrated in a small portion of the boundary, thus rendering the exit a narrow escape problem. Unlike the case in the classical activated escape problem, the principal eigenvalue does not necessarily decay exponentially as the relative noise strength decays. The oscillatory narrow escape problem arises in a mathematical model of neural networks with synaptic depression. We identify oscillation peaks in the density of the time the network spends in a specific state. This observation explains the oscillations of stochastic trajectories around a focus prior to escape and also the non-Poissonian escape times. This phenomenon has been observed and reported in the neural network literature.