Utilizing Generative Deep Learning for Variational Calculations of Nuclear Few-Body Systems

Quantum many-body calculations can be complex to solve analytically as well as time intensive and computationally expensive to perform numerically. This project explores the use of generative machine learning to perform quantum many-body calculations. Deep learning offers a method of more accurately approximating the wavefunction for such nuclear systems, allowing to calculate properties of atomic nuclei more efficiently. The Ferminet Neural Network framework was used to predict the wavefunction of the deuteron. Experimentation was made with network training in order to increase the accuracy and speed of convergence. In these pursuits, learning rate, batch size, epochs, hidden layers, and additional network hyperparameters were optimized for training.