Data for the publication "Performance of two complementary machine-learned potentials in modelling chemically complex systems", npj. Comp. Mat. This data set contains: the datasets of structures in cfg and npz formats, INCAR file which was used for VASP calculations, python script for reading npz format. These are essentially the 2-, 3-, and 4-component configurations (converted from OUTCARs) used to train families of machine-learning potentials. Data contains both 0K and finite-T structures of Ta-V-Cr-W subsystems, approx. 6000 configurations in total. The "in-distribution" data has 10 splits onto training/testing parts (in 80%/20% proportion), for the cross-validation tests. The öut-of-distribution" data is not split, it is used only for testing the accuracy.
Please log in
to take part in the discussion (add own reviews or comments).