1: \begin{abstract}

2:   We develop in this work a numerical method for stochastic

3:   differential equations (SDEs) with weak second order accuracy based

4:   on Gaussian mixture.  Unlike the conventional higher order schemes

5:   for SDEs based on It\^o-Taylor expansion and iterated It\^o

6:   integrals, the proposed scheme approximates the probability measure

7:   $\mu(X^{n+1}|X^n=x_n)$ by a mixture of Gaussians. The solution at

8:   next time step $X^{n+1}$ is drawn from the Gaussian mixture

9:   with complexity linear in dimension $d$. This provides a new

10:   general strategy to construct efficient high weak order numerical

11:   schemes for SDEs.

12:   \newline

13:   

14:   {{\bf Key words:} Gaussian mixture, stochastic differential equation, second order scheme, weak convergence}

15:   \newline

16:   

17:   {{\bf AMS subject classifications:} 60H35, 65C30, 65L20}

18: \end{abstract}

19: