GNN Models and results for the paper "Band-gap regression with architecture-optimized message-passing neural networks"

Contains files with model parameters for random search and reference models, as well as the converted AFLOW dataset, in graphs form. Corresponds to results in https://arxiv.org/pdf/2309.06348.pdf.

Model predictions along with AUID identifiers are located in result_combined.zip, band gap (egap) and formation energy (ef) predictions are from the PaiNN ensemble, band gap classification is done by MPEU model.

New results include PaiNN NAS models.

Compatible source code can be found at jraph_MPEU GitHub repository.