Using neural networks for enhanced sampling in computational biophysics
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
data/image_data
helper_func
CV
PlumedImagesNeuralNetwork.ipynb
PlumedPyTorchAutoEncoderCV.ipynb
PyTorchMetadynamics.ipynb
README.md
autoencoder.net
image_plumed.dat
imagenetwork.net
images.xtc
nework.net
plumed_autoencoder.dat

README.md

tf_metadynamics: Using PyTorch to enhance molecular simulations and using Plumed to classify images

This repo is a fun weekend project designed to show how complex PyTorch computational graphs can be turned into collective variables inside Plumed. This was done in two parts.

PyTorchMetadynamics.ipynb

This Jupyter notebook encodes Metadynamics into PyTorch using a custom loss function depenedent on the history. The forces then become the negative of the derivatives which are automatically obtained via back propagation. This was performed on the Muller potential.

PlumedImagesNeuralNetwork.ipynb

This Jupyter Notebook transfers a 3-layer Image Net PyTorch Classifier into Plumed, encodes the images as molecular trajectories, and use Plumed to predict the un-normalized image scores. The plumed input file image_plumed.dat has the actual plumed neural network script.