Software

SODAS++: A Python toolkit for performing unsupervised graph manifold projections of atomistic trajectories/structures.

Website: Now integrated as part of Catalyst (see below)

Aroboto, B., Chen, S., Hsu, T., Wood, B. C., Jiao, Y., Chapman, J., “Universal and interpretable classification of atomistic structural transitions via unsupervised graph learning“, Applied Physics Letters, September 2023 PDF

Catalyst: A Python API for atomic graph construction and graph neural network models (supervised and unsupervised).

Website: Materials-Informatics-Laboratory/Catalyst (github.com)