Abstract
We propose an accurate data-driven numerical scheme to solve StochasticDifferential Equations (SDEs), by taking large time steps. The SDEdiscretization is built up by means of a polynomial chaos expansion method, onthe basis of accurately determined stochastic collocation (SC) points. Byemploying an artificial neural network to learn these SC points, we can performMonte Carlo simulations with large time steps. Error analysis confirms thatthis data-driven scheme results in accurate SDE solutions in the sense ofstrong convergence, provided the learning methodology is robust and accurate.With a variant method called the compression-decompression collocation andinterpolation technique, we can drastically reduce the number of neural networkfunctions that have to be learned, so that computational speed is enhanced.Numerical results shows the high quality strong convergence error results, whenusing large time steps, and the novel scheme outperforms some classicalnumerical SDE discretizations. Some applications, here in financial optionvaluation, are also presented.