/data-driven-CVs

Supporting material for the paper "Data driven collective variables for enhanced sampling"

Primary LanguageBrainfuck

Data-driven collective variables for enhanced sampling

Luigi Bonati, Valerio Rizzi, and Michele Parrinello, J. Phys. Chem. Lett. 11, 2998-3004 (2020).

DOI arXiv plumID:20.004 MaterialsCloud

Important

This repository is kept as supporting material for the manuscript, but it is no longer updated. Check out the mlcolvar library for data-driven CVs, where you can find up-to-date tutorials and examples.

This repository contains:

  1. the code necessary to train and use a neural-network collective variable optimized with Fisher's discriminant analysis
  2. input file to reproduce the simulations reported in the paper

Requirements

  • Pytorch and LibTorch (v == 1.4)
  • PLUMED2

Tutorial

Here you can find a Google Colab notebook with the code and instructions for Deep-LDA CV training and export.

Code and results availability

The code and input files are available also on the PLUMED-NEST archive while the results of the simulations are available in the Materials Cloud repository.

Errata

There is a typo in the definition of \widetilde{\mathbf{w}}_i below eq. 7. The correct formula is \widetilde{\mathbf{w}}_i=\mathbf{L}^T \mathbf{w}_i .

Contact

If you have comments or questions please send an email to luigi bonati [at] phys chem ethz ch .