This tutorial uses a combination of slides and interactive jupyter notebooks using the
e3nn framework to present fundamental concepts about Euclidean equivariant neural networks: data types, equivariant operations, how symmetry effects simple tasks, and how to construct the building blocks of
e3nn. This tutorial was originally presented at IPAM at UCLA on November 14, 2019.
Convolutional neural networks for point clouds and voxels with Euclidean symmetries (3D translation and 3D rotation). The current code is a merger of the work from Tensor field networks and 3D Steerable CNNs.
Zhantao Chen, Nina Andrejevic, Nathan Drucker, Thanh Nguyen, R Patrick Xian, Tess Smidt, Yao Wang, Ralph Ernstorfer, Alan Tennant, Maria Chan, Mingda Li, arXiv (2021). arXiv:2102.03024
Simon Batzner, Tess E. Smidt, Lixin Sun, Jonathan P. Mailoa, Mordechai Kornbluth, Nicola Molinari, Boris Kozinsky, arXiv (2021). arXiv:2101.03164
Benjamin Kurt Miller, Mario Geiger, Tess E. Smidt, Frank Noé, arXiv (2020). arXiv:2008.08461
Maserati, L. et al, Materials Horizons (2020). DOI:10.1039/C9MH01917K
Smidt, T.E., Mack, S.A., Reyes-Lillo, S.E. et al. Sci Data 7, 72 (2020).
figshare data deposit