A new numerical method for 1-D backward stochastic differential equations without using conditional expectations (Q778246)

The paper starts by introducing a numerical scheme for a linear backward stochastic differential equation (BSDE). To deal with a nonlinear BSDE, a linear approximation of it is first obtained. Then the numerical scheme applied to the linear approximation produces approximations of the nonlinear BSDE. It is noted that no computation of conditional expectations is required. Data for an example is presented to demonstrate the accuracy of the approximations. An application of this approach is given wherein numerical solution of a class of quasi-linear deterministic partial differential equations is performed by identifying an associated BSDE and then using numerical approximations of the BSDE.