Data for "Robust recognition and exploratory analysis of crystal structures via Bayesian deep learning" (Nat. Commun., 2021)

Here, we provide all training and test data generated in the study "Robust recognition and exploratory analysis of crystal structures via Bayesian deep learning" (Nat. Commun., 2021). This data is provided via Atomic Simulation Environment (ASE) database files (*.db) as well as via *.tar.gz files, containing calculated Smooth-overlap-of-atomic-positions (SOAP) descriptors, geometry files and additional metadata. In addition, the model file (.h5) is provided where two json files specify the relation between output neuron index and class label