Master thesis defense by Alexandra Melitta Haslund-Gourley

Study of Neural Network Quantum States for Strongly Correlated Systems

Neural Network Quantum States (NNQS) offer a promising approach to solving strongly correlated systems in quantum chemistry by parameterizing the wavefunction directly as a neural network function of real-space electron coordinates, optimized variationally to minimize the ground-state energy. This Master's Thesis defense presents an investigation of the scaling behavior of NNQS as a function of system size and electron correlation. This is achieved through an analysis of the Markov Chain Monte Carlo sampling procedure and a hyperparameter optimization study. The talk will review the NNQS method, introduce the two experiments, and present preliminary scaling results, placing these predictions in comparison with existing computational methods on cost and accuracy.