Revisiting Multi-Permutation Equivariance through the Lens of irreducible Representations
Abstract
This paper explores the characterization of equivariant linear layers for representations of permutations and related groups. Unlike traditional approaches, which address these problems using parameter-sharing, we consider an alternative methodology based on irreducible representations and Schur's lemma. Using this methodology, we obtain an alternative derivation for existing models like DeepSets, 2-IGN graph equivariant networks, and Deep Weight Space (DWS) networks. The derivation for DWS networks is significantly simpler than that of previous results. Next, we extend our approach to unaligned symmetric sets, where equivariance to the wreath product of groups is required. Previous works have addressed this problem in a rather restrictive setting, in which almost all wreath equivariant layers are Siamese. In contrast, we give a full characterization of layers in this case and show that there is a vast number of additional non-Siamese layers in some settings. We also show empirically that these additional non-Siamese layers can improve performance in tasks like graph anomaly detection, weight space alignment, and learning Wasserstein distances. Our code is available at GitHub.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
1 top-tier paper has cited it since. Know what they found before your reviewers do.
Lune can draw on
the full paper · the top-tier papers that cite it · related work
Free to start. No credit card required.
Lineage
- Equivariant Subgraph Aggregation Networks cites On Learning Sets of Symmetric Elements
- A Practical Method for Constructing Equivariant Multilayer Perceptrons for Arbitrary Matrix Groups cites On Learning Sets of Symmetric Elements
- Equivariant Architectures for Learning in Deep Weight Spaces cites On Learning Sets of Symmetric Elements
- Equivariant Architectures for Learning in Deep Weight Spaces cites A Practical Method for Constructing Equivariant Multilayer Perceptrons for Arbitrary Matrix Groups
- Equivariant Architectures for Learning in Deep Weight Spaces cites Optimizing Mode Connectivity via Neuron Alignment
- Permutation Equivariant Neural Networks for Symmetric Tensors cites On Learning Sets of Symmetric Elements
- Permutation Equivariant Neural Networks for Symmetric Tensors cites A Practical Method for Constructing Equivariant Multilayer Perceptrons for Arbitrary Matrix Groups
Cited by top-tier papers1
Builds on20
- A Practical Method for Constructing Equivariant Multilayer Perceptrons for Arbitrary Matrix GroupsMarc Finzi, Max Welling, Andrew Gordon WilsonICML 2021 · 226 citations
- Equivariant Subgraph Aggregation NetworksBeatrice Bevilacqua, Fabrizio Frasca, Derek Lim, Balasubramaniam Srinivasan et al.ICLR 2022 · 217 citations
- On Learning Sets of Symmetric ElementsHaggai Maron, Or Litany, Gal Chechik, Ethan FetayaICML 2020 · 148 citations
- Optimizing Mode Connectivity via Neuron AlignmentN. Joseph Tatro, Pin-Yu Chen, Payel Das, Igor Melnyk et al.NeurIPS 2020 · 104 citations
- Equivariant Architectures for Learning in Deep Weight SpacesAviv Navon, Aviv Shamsian, Idan Achituve, Ethan Fetaya et al.ICML 2023 · 101 citations
Related papers
- On Universality Classes of Equivariant NetworksMarco Pacini, Gabriele Santin, Bruno Lepri, Shubhendu TrivediNeurIPS 2025 · 4 citations
- Graph Metanetworks for Processing Diverse Neural ArchitecturesDerek Lim, Haggai Maron, Marc T. Law, Jonathan Lorraine et al.ICLR 2024 · 47 citations
- Any-Subgroup Equivariant Networks via Symmetry BreakingAbhinav Goel, Derek Lim, Hannah Lawrence, Stefanie Jegelka et al.ICLR 2026
- Identifiable Equivariant Networks are Layerwise EquivariantVahid Shahverdi, Giovanni Luca Marchetti, Georg Bökman, Kathlén KohnICML 2026
- Understanding and Extending Subgraph GNNs by Rethinking Their SymmetriesFabrizio Frasca, Beatrice Bevilacqua, Michael M. Bronstein, Haggai MaronNeurIPS 2022 · 168 citations
