This is a repository for the final project of the course Computational Statistical Physics (PHYS 7810) at CU Boulder, of which the goal is to develop and apply Boltzmann generators to different molecular systems. The project paper can be found in the folder Project. Specifically, this repository includes the following directories:
Project: A folder containing the project paper, presentation slides, and pictures for the final proejct.References: A folder containing the most important journal papers relevant to our project.Notebooks: A folder containing severl jupyter notebooks in a tutorial style which implement Boltzmann generators and applies them to different systems of interest. From the introductory tutorial of PyTorch and the simplest toy model, to a more advanced molecular system, these notebooks include:
PyTorch Introduction.ipynb: A jupyter notebook adapted from a tutorial generously provided by the course CSE446: Machine Learning at Univeristy of Washington.Double-well Potential.ipynb: A notebook which implements the architecture of Boltzmann generators and applies them to the simplest toy model: double-well potential.Mueller Brown Potential.ipynb: A notebook which applies Boltzmann generators to the slightly more complex Mueller potential, which is characteristic of three energy minima and a more complicated reaction coordinate.Dimer-Simulation.ipynb: A notebook which applies Boltzmann generators to a dimer in Lennard-Jones bath.Library: A folder of software implementing Boltzmann generators that can be imported by the jupyter notebooks in the folder Notebooks.Both authors contributed equally to the project.
Wei-Tse Hsu
generator.py, training.py, density estimator.py, visuals.py.)Double-well Potential.ipynb).Lenny Fobe
Mueller Brown Potential.ipynb, Dimer-Simulation.ipynb).Copyright (c) 2019