Multilevel Monte Carlo methods

With Monte Carlo methods, to achieve improved accuracy one often requires more expensive sampling (such as a finer timestep discretisation of a stochastic differential equation) in addition to more samples. Multilevel Monte Carlo methods aim to avoid this by combining simulations with different levels of accuracy. In the best cases, the average cost of each sample is independent of the overall target accuracy, leading to very large computational savings. The talk will emphasise the simplicity of the approach, give an overview of the range of applications being worked on by various researchers, and mention some recent extensions including work by Peter Glynn and Chang-han Rhee. Applications to be discussed will include financial modelling, engineering uncertainty quantification, stochastic chemical reactions, and the Feynman-Kac formula for high-dimensional parabolic PDEs. Further information can be obtained from http://people.maths.ox.ac.uk/gilesm/acta/Non UBCUnreviewedAuthor affiliation: Oxford UniversityFacult