andim53 / crystalfeatures

CrystalFeatures

CrystalFeatures is a set of python codes that extract features of crystal systems, using only the CIF file as input, for machine learning applications. There are three classes of features in CrystalFeatures: basic atomic and crystal descriptors (BACD), which include statistical properties of the atoms in the materials as well as the crystal symmetry; geometry features, which are based on bonding properties; and robust one-shot ab initio (ROSA) descriptors, where the eigen values of the material are obtained using a single-point single step calculation is performed using GPAW.

Data for ROSA paper

Link to paper: https://jcheminf.biomedcentral.com/articles/10.1186/s13321-022-00658-9

In order to reproduce the results in the manuscript, the trained XGBOOST models are provided in the folder /data/models/, and the test set data are provided in the folder /data/test/ for each of the properties listed in the paper.

Please note

To generate ROSA descriptors using GPAW, you need to modify the scf.py code of your GPAW according to the one in this repository. If you are using GPAW version 21.1.0, you can simply overwrite your scf.py with the one provided here. Otherwise, just comment a few lines of code in your scf.py as shown in the scf.py provided here. To find those lines of code: download scf.py from this repository and look up "SHERIF". I have added comments at the lines of code that need commenting in your scf.py.