Title:Efficient Bayesian estimation of the generalized Langevin equation from data

Abstract:The generalized Langevin equation (GLE) overcomes the limiting Markov approximation of the standard Langevin equation (SLE) by an incorporated memory kernel and can be used to model various stochastic processes in many fields of science ranging from climate modeling over neuroscience to finance. Generally, Bayesian estimation facilitates the data-driven determination of both suitable model parameters and their credibility in a straightforward way. In this work, we discuss this approach to estimating the generalized Langevin equation in the case of white noise, arbitrary drift and diffusion functions, and an arbitrary time-discrete kernel function with a fixed length $K$. We show how a reasonable value of $K$ can be determined based on the data. Moreover, we propose a modification of the memory term, leading to an equivalent model with an emphasis on the impact of trends, which ensures that an estimate of the SLE provides an effective initial guess for the GLE. We further improve the efficiency of the procedure through a piecewise constant approximation of the drift and diffusion functions. In this case, the characteristics of the data are represented by only a few coefficients, which allows the analysis of even large data sets. We illustrate the capabilities of both the method and the model by an example from turbulence.