Many settings in machine learning require the selection of a rotation representation. However, choosing a suitable representation from the many available options is challenging. This paper acts as a survey and guide through rotation representations. We walk through their properties that harm or benefit deep learning with gradient-based optimization. By consolidating insights from rotation-based learning, we provide a comprehensive overview of learning functions with rotation representations. We provide guidance on selecting representations based on whether rotations are in the model's input or output and whether the data primarily comprises small angles.
| Author(s): | Geist, Andreas René and Frey, Jonas and Zhobro, Mikel and Levina, Anna and Martius, Georg |
| Book Title: | Proceedings of the Forty-First International Conference on Machine Learning |
| Journal: | Proceedings of Machine Learning Research |
| Volume: | 235 |
| Pages: | 15331--15350 |
| Year: | 2024 |
| Month: | July |
| Editors: | Salakhutdinov, Ruslan and Kolter, Zico and Heller, Katherine and Weller, Adrian and Oliver, Nuria and Scarlett, Jonathan and Berkenkamp, Felix |
| Project(s): | |
| Bibtex Type: | Conference Paper (inproceedings) |
| Event Name: | Forty-First International Conference on Machine Learning |
| Event Place: | Vienna, Austria |
| State: | Published |
| URL: | https://proceedings.mlr.press/v235/geist24a.html |
| Electronic Archiving: | grant_archive |
| Organization: | ICML |
BibTex
@inproceedings{Geistetal2024:rotations,
title = {Learning with 3D rotations, a hitchhiker's guide to SO(3)},
journal = {Proceedings of Machine Learning Research},
booktitle = {Proceedings of the Forty-First International Conference on Machine Learning },
abstract = {Many settings in machine learning require the selection of a rotation representation. However, choosing a suitable representation from the many available options is challenging. This paper acts as a survey and guide through rotation representations. We walk through their properties that harm or benefit deep learning with gradient-based optimization. By consolidating insights from rotation-based learning, we provide a comprehensive overview of learning functions with rotation representations. We provide guidance on selecting representations based on whether rotations are in the model's input or output and whether the data primarily comprises small angles.},
volume = {235},
pages = {15331--15350},
editors = {Salakhutdinov, Ruslan and Kolter, Zico and Heller, Katherine and Weller, Adrian and Oliver, Nuria and Scarlett, Jonathan and Berkenkamp, Felix},
organization = {ICML},
month = jul,
year = {2024},
slug = {geistetal2024-rotations},
author = {Geist, Andreas Ren\'{e} and Frey, Jonas and Zhobro, Mikel and Levina, Anna and Martius, Georg},
url = {https://proceedings.mlr.press/v235/geist24a.html},
month_numeric = {7}
}