2017/01/01 by Sean Carnaffan, Reiichiro Kawai
Mathematics · Physics and Astronomy · #Fractional Differential Equations Solutions #Statistical Distribution Estimation and Applications #stochastic dynamics and bifurcation
paper · doi:10.1137/17m111482x
openalex publication_date 2017/01/01 · openalex created_date 2025/10/10 · openalex updated_date 2026/06/11
Fractional Fokker--Planck equations are generalizations of ordinary Fokker--Planck equations that include fractional derivative operators to encapsulate trapping effects in the anomalously diffusing population they describe. They do not admit analytical solutions in general and there are significant drawbacks to current numerical approximation methods. Using Malliavin calculus, we establish formulas in the form of a mathematical expectation for the probability density functions of anomalous diffusion processes, which are known to solve fractional Fokker--Planck equations. These density formulas lend themselves easily to computation via Monte Carlo sampling and are easy to implement. Moreover, they have the capability to deal with a general model of anomalous diffusion, including the case where the parent process of the anomalous diffusion has a Lévy jump component. The numerical solutions obtained are inherently unbiased, which allows them to form the basis for reliable inferences about the population dynamics. We outline an easy-to-use implementation scheme including methods for generating sample paths of anomalous diffusion processes driven by the inverse stable, inverse tempered stable, and inverse gamma subordinators. We illustrate the effectiveness and accuracy of our method and the physical insight it brings in both univariate and multivariate settings.