https://github.com/VincentStimper/hmc-hyperparameter-tuning

Framework to tune the hyperparameters of Hamiltonian Monte Carlo in an automated fashion
https://github.com/VincentStimper/hmc-hyperparameter-tuning

Last synced: 3 months ago
JSON representation

Framework to tune the hyperparameters of Hamiltonian Monte Carlo in an automated fashion

• Host: GitHub
• URL: https://github.com/VincentStimper/hmc-hyperparameter-tuning
• Owner: VincentStimper
• License: mit
• Created: 2020-10-01T17:36:31.000Z (about 4 years ago)
• Default Branch: main
• Last Pushed: 2022-08-10T11:54:39.000Z (over 2 years ago)
• Last Synced: 2024-07-04T00:56:43.640Z (4 months ago)
• Language: Python
• Homepage: https://proceedings.mlr.press/v139/campbell21a.html
• Size: 201 KB
• Stars: 3
• Watchers: 2
• Forks: 2
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Tuning hyperparameters of Hamiltonian Monte Carlo

This is the code used to do the experiments of the article [A Gradient Based Strategy for Hamiltonian Monte Carlo

Hyperparameter Optimization](https://proceedings.mlr.press/v139/campbell21a.html). 

## Installation instructions

### 2D toy examples

These experiments were done with numpy, scipy, and autograd, so these packages have to be installed.

### Variational Autoencoder

The framework TensorFlow 1.15 was used here. For installation instruction, see https://www.tensorflow.org/install.

### 1D experiments and molecular configuration sampling

We used the framework PyTorch 1.6 for these experiments, see  https://pytorch.org/get-started/locally/ for installation

instructions. The experiments involving alanine dipeptide require [OpenMM](https://openmm.org/) to be

installed, which can be done via [conda](https://anaconda.org/conda-forge/openmm). The other dependencies

can be installed via

```

pip install -r requirements.txt

```

The scripts for running the experiments are in the `molecular-configurations` directory. Each experiment can be

reproduced using the respective configuration file in `molecular-configurations/config`.