Runge-Kutta methods for third order weak approximation of SDEs with multidimensional additive noise (Q1960213)

The author constructs a new class of third order Runge-Kutta numerical methods for stochastic differential equations with additive noise. The order conditions up to order three are given in the paper, and a particular explicit third-order method is derived. The author compares the method to some well known second order methods and to the only third order method for weak approximation he knows, Platen's method [cf. \textit{P. E. Kloeden, E. Platen}, Numerical solution of stochastic differential equations. Applications of Mathematics. 23. Berlin: Springer-Verlag. (1992; Zbl 0752.60043)]. The comparison shows that his new method seem give to give better results in terms of error and computational efforts. It is regretful, however, that the author does not compare his method to a recent methods for weak approximation of stochastic differential equations based on cubature on Wiener spaces, as explained e.g. in [\textit{S. Ninomiya} and \textit{N. Victoir}, Appl. Math. Finance 15, No. 2, 107--121 (2008; Zbl 1134.91524)]. The latter methods proved very efficient and work for a general, not necessarily additive, noise.