(192bk) Machine Learning Approaches to Quantum Monte Carlo Challenges

Quantum Monte Carlo (QMC) methods provide an accurate and highly parallelisable computational approach to quantum chemical simulations of molecules. However, challenges still exist in the development of QMC algorithms for improved efficiency and accuracy, as well as calculation of properties beyond molecular energies. For example, the optimisation of molecular geometries with QMC amounts to finding the minimum of a noisy surface - a tricky but common problem. We are investigating the development of machine learning algorithms for application to such QMC challenges. Effective implementation of these machine learning techniques allows for accurate and efficient calculation of quantum-scale molecular properties.