Direct Molecular Polarizability Prediction with ๐บ โ ๐ถ โ ( ๐ ) \boldsymbol{SO(3)} Equivariant Local Frame GNNs
Jean Philip Filling, Felix Post, Michael Wand, Denis Andrienko ยท 2025
Read Direct Molecular Polarizability Prediction with ๐บ โ ๐ถ โ ( ๐ ) \boldsymbol{SO(3)} Equivariant Local Frame GNNs by Jean Philip Filling, Felix Post, Michael Wand, Denis Andrienko (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-07087.txt.
Description
Jean Philip Filling, Felix Post, Michael Wand, and Denis Andrienko design a neural network that predicts molecular polarizability as a tensor. Polarizability describes how a molecule's electron cloud distorts in an electric field, and it is inherently directional. Standard models regress scalar quantities and recover tensor parts only through derivatives, losing geometric structure along the way.
Their graph neural network maintains rotational equivariance by working in local coordinate frames attached to the molecule. Scalar, vector, and tensor channels all pass messages within this local system, so the geometry is respected at every step. On the QM7-X dataset, tensorial message passing beat scalar message passing models at predicting polarizabilities. The authors present the design as a step toward structured, geometry-aware learning for molecular response.
Catalog record
| CREATOR | Jean Philip Filling, Felix Post, Michael Wand, Denis Andrienko |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 21 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-07087