Random Fourier feature approximations of potential in Hamiltonian systems yield approximations of molecular dynamics. Particularly in this small-compute project, we are interested in studying the convergence rate of our random feature algorithm with the increasing number of nodes in the network and the increasing size of the training data set. Specifically we plan to use Monte Carlo method to compute the correlation function between observables, obtained by the molecular dynamics results for a Hamiltonian system using the trained potential form the random feature representation. In order to achieve good accuracy by the Monte Carlo computation, we will utilize the parallel functionality of Matlab for generating a sufficiently large training data set and also for validating our model. We also aim to use parallelization in the training process with Python and Tensorflow. In the continuation of the proposed project, we mainly aim to implement more empirical tests so as to provide numerical verifications under various conditions, to satisfy the suggested improvements by our referees.