On the pathwise approximation of stochastic differential equations (Q329029)

Non-traditional assumptions are identified and used to prove pathwise convergence when numerically approximating the solution of \(d\)-dimensional Ito stochastic differential equations. These assumptions are verified to prove convergence of the fixed time-stepping and two new adaptive time-stepping versions of the Euler-Maruyama method. Tighter error bounds are established for the two adaptive versions that facilitate error control when these methods are implemented. For some examples numerical results are presented that indicate that the adaptive methods can yield smaller error with less variation but the required sampling of bounded diffusion may make them more expensive.