Abstract

In this talk, we introduce a deep learning-based efficient approximation algorithm for a high-dimensional nonlinear partial differential equation (PDE) which typically appears in credit valuation adjustment (CVA) computation. While it is difficult for the classical numerical methods for nonlinear PDEs to implement for high-dimensional models due to their computational costs, deep learning-based methods work even when the dimension of the PDE is high. Combining deep learning and high-order discretization, the proposed method achieves an efficient approximation for CVA computation in high-dimensional settings. Numerical experiments confirm the accuracy of the proposed method. This talk is based on a joint work with Toshihiro Yamada.