1: \begin{abstract}

2: %	Recently, lots of researches about solving high-dimensional backward stochastic differential equations (BSDEs) by deep learning method has appeared. 

3: {\color{vert}

4: \textcolor{black}{We propose a new probabilistic scheme which combines deep learning techniques with high order schemes for backward stochastic differential equations belonging to the class of Runge-Kutta methods to solve high-dimensional semi-linear parabolic partial differential equations.} Our approach notably extends the one introduced in \cite{hure2020deep} for the implicit Euler scheme to schemes which are more efficient in terms of discrete-time error. We establish some convergence results for our implemented schemes under classical regularity assumptions. We also illustrate the efficiency of our method for different schemes of order one, two and three. Our numerical results indicate that the Crank-Nicolson schemes is a good compromise in terms of precision, computational cost and numerical implementation.

5: %

6: %	Deep learning techniques are efficient technique to overcome empirically the curse of dimensionality for solving high-dimensional backward stochastic differential equations (BSDEs). 

7: %The current deep learning algorithms are based on an Euler type  discretization. In this work, we  instead combine, some high-order  time discretization schemes  such as Crank-Nicolson scheme or explicit multi stage Runge-Kutta scheme with non-linear regression. We prove theoretical convergence bounds for our algorithms. We then numerically compare the computational time cost of different methods and show that high order scheme for the discrete time error, if correctly implemented, are more efficient than classical Euler schemes.

8: }

9: \end{abstract}

10: