Application of data driven models in quantum mechanics calculations

Minimization of the total potential energy of a system of interacting atoms is a classical challenging problem in physics. Standard quantum mechanics calculation of the total potential energy is a very time consuming task. The article reports preliminary results of our new investigation of using data driven models as a means of approximating the total system energy. Such models include feedforward artificial neural networks and linear least squares models using radial basis functions and polynomial basis functions. So far, we have considered a system of two atoms and used the two-body Lenard-Jones potential to generate training data. We have found several types of basis functions that are suitable for the approximation.