Tensorflow implementation of message passing neural networks for molecules and materials. The framework implements the SchNet model and its extension with edge update network NMP-EDGE as well as the model used in Materials property prediction using symmetry-labeled graphs as atomic-position independent descriptors. The main difference between msgnet and schnetpack is that msgnet follows a message passing architecture and can therefore be more flexible in some cases, e.g. it can be used to train on graphs rather than on structures with full spatial information.
Jørgensen, Peter Bjørn, et al. "Materials property prediction using symmetry-labeled graphs as atomic position independent descriptors." Physical Review B 100.10 (2019): 104114.
message passing neural networks, Tensorflow, symmetry-labeled graphs,
Curator: qinyang