English

Training Dynamics of Learning 3D-Rotational Equivariance

Machine Learning 2025-12-03 v1 Biomolecules

Abstract

While data augmentation is widely used to train symmetry-agnostic models, it remains unclear how quickly and effectively they learn to respect symmetries. We investigate this by deriving a principled measure of equivariance error that, for convex losses, calculates the percent of total loss attributable to imperfections in learned symmetry. We focus our empirical investigation to 3D-rotation equivariance on high-dimensional molecular tasks (flow matching, force field prediction, denoising voxels) and find that models reduce equivariance error quickly to \leq2\% held-out loss within 1k-10k training steps, a result robust to model and dataset size. This happens because learning 3D-rotational equivariance is an easier learning task, with a smoother and better-conditioned loss landscape, than the main prediction task. For 3D rotations, the loss penalty for non-equivariant models is small throughout training, so they may achieve lower test loss than equivariant models per GPU-hour unless the equivariant ``efficiency gap'' is narrowed. We also experimentally and theoretically investigate the relationships between relative equivariance error, learning gradients, and model parameters.

Keywords

Cite

@article{arxiv.2512.02303,
  title  = {Training Dynamics of Learning 3D-Rotational Equivariance},
  author = {Max W. Shen and Ewa Nowara and Michael Maser and Kyunghyun Cho},
  journal= {arXiv preprint arXiv:2512.02303},
  year   = {2025}
}

Comments

Accepted to Transactions on Machine Learning Research (TMLR)

R2 v1 2026-07-01T08:04:51.784Z