Poster
in
Workshop: Symmetry and Geometry in Neural Representations (NeurReps)
Equivariant Representations for Non-Free Group Actions
Luis Armando Pérez Rey · Giovanni Luca Marchetti · Danica Kragic · Dmitri Jarnikov · Mike Holenderski
Keywords: [ Representation Learning ] [ Group Theory ] [ symmetries ]
We introduce a method for learning representations that are equivariant with respect to general group actions over data. Differently from existing equivariant representation learners, our method is suitable for actions that are not free i.e., that stabilize data via nontrivial symmetries. Our method is grounded in the orbit-stabilizer theorem from group theory, which guarantees that an ideal learner infers an isomorphic representation. Finally, we provide an empirical investigation on image datasets with rotational symmetries and show that taking stabilizers into account improves the quality of the representations.