Multiplicative cascades applied to PDEs (two numerical examples) (Q2490276): Difference between revisions

The aim of this paper is to propose techniques of Monte Carlo simulation of certain random multiplicative cascades in order to obtain improved numerical approximations of the Fourier transformed solution of partial differential equations (PDEs). Stochastic processes have been found to have important connections to deterministic PDEs, the solution to the PDE being represented as the expected value of a functional acting on the sample paths of a certain stochastic process evolving in the physical space. This paper deals with the analogous idea in the Fourier space, multiplicative functionals of tree-like stochastic models being used to give probabilistic representation of the Fourier transform of the PDEs solution. Two particular equations are considered: a linear diffusion equation and a viscous Burgers equation. Further restrictions are imposed to the Fourier transform of the data in each PDE to achieve the probabilistic representation. For the particular analyzed equations, the transition distributions for the frequency process are chosen such that both the analytical and the modelling problem to be considerably simplified. This selection imposes restrictions on the PDE data, for which a multiplicative cascade representation gives a solution. The main emphasis of the paper is on the design of Monte Carlo simulation schemes to numerically approximate the solution of each PDE in the Fourier space. The algorithm proposed exploit the structure of the branching random walks on which multiplicative cascades are defined. The computing results show initial numerical approximations with errors less than 5\% in the leading Fourier coefficients of the solution. This approximation is further improved by means of a (fixed-point) Picard iteration scheme on the integral equation with representation of PDEs in the Fourier space.