Code for our NeurIPS 2025 Oral paper Learning (Approximately) Equivariant Networks via Constrained Optimization.

Code cleanup in progress, all code and instructions for reproducing experiments will be make publicly available soon.

Status:

• SEGNN + N-Body
• SEGNN + QM9
• SchNet + QM9
• EGNO + MoCap
• VN-DGCNN + ModelNet40

Please also refer to Appendix A.1. for an example implementation of ACE.

If you use the code from this repo, please consider citing us as:

@inproceedings{
manolache2025learning,
title={Learning (Approximately) Equivariant Networks via Constrained Optimization},
author={Andrei Manolache and Luiz F. O. Chamon and Mathias Niepert},
booktitle={The Thirty-ninth Annual Conference on Neural Information Processing Systems},
year={2025},
url={https://openreview.net/forum?id=NM4emKloy6}
}

Moreover, please consider citing the original papers for the base models.