Multilevel Monte Carlo for the Feynman–Kac formula for the Laplace equation
- 549 Downloads
Since its formulation in the late 1940s, the Feynman–Kac formula has proven to be an effective tool for both theoretical reformulations and practical simulations of differential equations. The link it establishes between such equations and stochastic processes can be exploited to develop Monte Carlo sampling methods that are effective, especially in high dimensions. There exist many techniques of improving standard Monte Carlo sampling methods, a relatively new development being the so-called Multilevel Monte Carlo method. This paper investigates the applicability of multilevel ideas to the stochastic representation of partial differential equations by the Feynman–Kac formula, using the Walk on Spheres algorithm to generate the required random paths. We focus on the Laplace equation, the simplest elliptic PDE, while mentioning some extension possibilities.
KeywordsMultilevel Monte Carlo Feynman–Kac Walk on Spheres Laplace equation
Mathematics Subject Classification60H30 65C05 65N99
We thank the anonymous reviewers for their valuable comments and suggestions that improved the quality of the paper.
- 4.Binder, I., Braverman, M.: The complexity of simulating Brownian motion. In: Proceedings of the 20th annual ACM-SIAM symposium on discrete algorithms, pp. 58–67 (2009)Google Scholar
- 5.Bruijn, N.: Asymptotic Methods in Analysis. Bibliotheca mathematica. Dover Publications, Mineola (1970)Google Scholar
- 7.Chang, J.: Stochastic Processes. http://www.stat.yale.edu/jtc5/251/stochastic-processes . Lecture notes (2007)
- 9.Gantner, R. N.: Computing the Feynman–Kac formula efficiently with multilevel Monte Carlo. Master thesis, Computer Science Department, ETH Zurich (2013)Google Scholar
- 17.Nevanlinna, R.H.: Analytic Functions. Princeton Mathematical Series. Princeton University Press, Princeton (1960)Google Scholar