This was a project from an introductory course in CUDA programming at the Sorbonne University, Paris. We implemented the nested Monte-Carlo parallelly in order to price a bullet option at every triplet (t,x,j).
Furthermore, we trained a linear regression and a neural network for interpolating the value function F(t,x,j).
The results are presented in the beamer presentation. In the end we compared the accuracy and execution time of three models and visualized the results in the following plot: