Timezone: »
SE(3)-equivariant self-attention via invariant features
Nan Chen · Soledad Villar
In this work, we use classical invariant theory to construct a self-attention module equivariant to 3D rotations and translations. The parameterization is based on the characterization of SE(3)-equivariant functions via the invariants ---scalar products of vectors and certain subdeterminants. This parameterization can be seen as a natural extension to a (more straightforward) E(3) equivariant attention based on invariants ---scalar products or pairwise distances of vectors. We evaluate our model using a toy N-body particle simulation dataset and a real-world dataset of molecular properties. Our model is easy to implement and it exhibits comparable performance and running time to state-of-the-art methods.
Author Information
Nan Chen (National University of Singapore)
Soledad Villar (Johns Hopkins University)
More from the Same Authors
-
2021 : A simple equivariant machine learning method for dynamics based on scalars »
Weichi Yao · Kate Storey-Fisher · David W Hogg · Soledad Villar -
2021 : Constraints with Doug Burger, Alysson Muotri, Ralph-Etienne-Cummings, Florian Engert »
Doug Burger · Florian Engert · Ralph Etienne-Cummings · Soledad Villar · Teresa Huang -
2022 : From Local to Global: Spectral-Inspired Graph Neural Networks »
Ningyuan Huang · Soledad Villar · Carey E Priebe · Da Zheng · Chengyue Huang · Lin Yang · Vladimir Braverman -
2022 : Keynote: Soledad Villar »
Soledad Villar -
2021 : Introduction »
Weiwei Yang · Joshua T Vogelstein · Onyema Osuagwu · Soledad Villar · Johnathan Flowers · Weishung Liu · Ronan Perry · Kaleab Alemayehu Kinfu · Teresa Huang