A multilevel Monte Carlo algorithm for Lévy-driven stochastic differential equations (Q550167)

The multilevel Monte Carlo method due to \textit{M. B. Giles} [Oper. Res. 56, No.~3, 607--617 (2008; Zbl 1167.65316)] relies on simulating a hierarchy of coupled Euler schemes with increasingly finer discretisation. Compared to coarse levels, fewer simulations are performed on the computationally demanding fine levels. This approach is transfered here to solutions to stochastic differential equations driven by Lévy processes. The discontinuous part of the Lévy process is dicretized by removing jumps below a certain threshold. The paper studies convergence of the corresponding multilevel algorithm. Whereas the rate of convergence generally depends on the Blumenthal-Getoor index, the optimal rate of order \(1/\sqrt{n}\) can be achieved for Blumenthal-Getoor indices below one.