@inproceedings{6591ab0fe0954925960e6dbdfc58f9bb,
title = "Equivariant Representation Learning in the Presence of Stabilizers",
abstract = "We introduce Equivariant Isomorphic Networks (EquIN) – a method for learning representations that are equivariant with respect to general group actions over data. Differently from existing equivariant representation learners, EquIN is suitable for group actions that are not free, i.e., that stabilize data via nontrivial symmetries. EquIN is theoretically grounded in the orbit-stabilizer theorem from group theory. This guarantees that an ideal learner infers isomorphic representations while trained on equivariance alone and thus fully extracts the geometric structure of data. We provide an empirical investigation on image datasets with rotational symmetries and show that taking stabilizers into account improves the quality of the representations.",
keywords = "Representation Learning, Group Theory, Equivariance, Lie Groups",
author = "\{Perez Rey\}, \{Luis A.\} and Marchetti, \{Giovanni Luca\} and Danica Kragic and Jarnikov, \{Dmitri S.\} and Holenderski, \{Mike J.\}",
year = "2023",
month = sep,
day = "18",
doi = "10.1007/978-3-031-43421-1\_41",
language = "English",
isbn = "978-3-031-43420-4",
volume = "IV",
series = "Lecture Notes in Computer Science (LNCS)",
publisher = "Springer",
pages = "693--708",
editor = "Danai Koutra and Claudia Plant and \{Gomez Rodriguez\}, Manuel and Elena Baralis and Francesco Bonchi",
booktitle = "Machine Learning and Knowledge Discovery in Databases: Research Track",
address = "Germany",
}