/GP-representations

Primary LanguagePythonGNU General Public License v3.0GPL-3.0

GP-representations

Scripts used to generate group and period-based Coordination Atom Fingerprints and FCHL19 representations (GP-CAF and GP-FCHL19).

As this is an improvement based on Coordination Atom Fingerprints and FCHL19.

It contains lots of code from the repos of above publications, which are

GASpy, GASpy_regressions, qml

As well as the repo used to get the dataset

CatHub

Installation

Requirements

1. Python environment

We recommend using Conda package manager

conda create -n gprep python=3.7.3
source activate gprep

2. Essential python packages

  • ase=3.18.1
  • pymatgen=2019.11.11
  • scikit-learn=0.22.1
  • tpot=0.11.0
  • xgboost=0.90
  • mendeleev==0.5.1 Thes can be installed through
conda env update --file gprep.yml

3. Installing QML, CatHub

As the GP-FCHL19 based on the Fortran/Python interface used in qml, qml is essential to the installation of this code. Please go to http://www.qmlcode.org/installation.html to get more information about QML installation. If QML cannot be installed, this code cannot as well.

Install QML

git clone -b develop https://github.com/qmlcode/qml.git
cd qml
python setup.py install
cd ../

Install CatHub

https://github.com/SUNCAT-Center/CatHub.git
cd CatHub
pip install .
cd ../

4. Installing GPREP

python setup.py install

Usage

python get_gpcaf.py
python get_gpfchl19.py

Acknowledgements

  • The GP-CAF is based on the CAF implemented in GASpy_regressions.
  • The GP-FCHL19 is based on the FCHL19 implemented in qml.