Skip to content

A repository for the final project implementing/applying Boltzmann generators for Computational Statistical Physics (PHYS 7810) at CU Boulder

Notifications You must be signed in to change notification settings

wehs7661/boltzmann_generators

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

76 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Boltzmann Generators

Description

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.

Contributions

  • Both authors contributed equally to the project.

  • Wei-Tse Hsu

    • Developed codes for performing Monte Carlo simulations.
    • Developed codes for building and training a Boltzmann generators and corresponding data analysis. (generator.py, training.py, density estimator.py, visuals.py.)
    • Applied Boltzmann generators to the double-well potential (Double-well Potential.ipynb).
    • Presentation slides (page 1 to page 20).
    • Project paper: sections relevant to the work above.
  • Lenny Fobe

    • Developed codes for performing molecular dynamics simulation.
    • Applied Boltzmann generators to the Muller Brown potential and the dimer in the Lennard-Jones bath. (Mueller Brown Potential.ipynb, Dimer-Simulation.ipynb).
    • Presentation slides (page 21 to page 37).
    • Project paper: sections relevant to the work above.

Authors

Copyright (c) 2019