mlsanityData to 14 Oct 2026

RECON: Robust symmetry discovery via Explicit Canonical Orientation Normalization

arXiv 2505.13289 ↗First posted 19 May 2025cs.LG cs.CVcited by 1

Not on the map. Why?

Abstract (arXiv)

Real world data often exhibits unknown, instance-specific symmetries that rarely exactly match a transformation group G fixed a priori. Class-pose decompositions aim to create disentangled representations by factoring inputs into invariant features and a pose g∈ G defined relative to a training-dependent, arbitrary canonical representation. We introduce RECON, a class-pose agnostic canonical orientation normalization that corrects arbitrary canonicals via a simple right translation, yielding natural, data-aligned canonicalizations. This enables (i) unsupervised discovery of instance-specific pose distributions, (ii) detection of out-of-distribution poses and (iii) a plug-and-play test-time canonicalization layer. This layer can be attached on top of any pre-trained model to infuse group invariance, improving its performance without retraining. We validate on images and molecular ensembles, demonstrating accurate symmetry discovery, and matching or outperforming other canonicalizations in downstream classification.

Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 54 of its 74 references that are in the map. First posted on arXiv 19 May 2025.

Builds on

54 of 74 references.

References · 54

  1. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,177
  2. 2021Emerging Properties in Self-Supervised Vision Transformerscited by 6,968
  3. 2020Bootstrap your own latent: A new approach to self-supervised Learningcited by 4,750
  4. 2019Fast Graph Representation Learning with PyTorch Geometriccited by 2,434
  5. 2015Spatial Transformer Networkscited by 1,429
  6. 2021Geometric Deep Learning: Grids, Groups, Graphs, Geodesics, and Gaugescited by 1,167
  7. 2016Group Equivariant Convolutional Networkscited by 1,165
  8. 2018Tensor field networks: Rotation- and translation-equivariant neural networks for 3D point cloudscited by 746
  9. 2019General E(2)-Equivariant Steerable CNNscited by 446
  10. 20183D Steerable CNNs: Learning Rotationally Equivariant Features in Volumetric Datacited by 354
  11. 2020Self-supervised Equivariant Attention Mechanism for Weakly Supervised Semantic Segmentationcited by 300
  12. 2018Towards a Definition of Disentangled Representationscited by 296
38 more
  1. 2019Gauge Equivariant Convolutional Networks and the Icosahedral CNNcited by 267
  2. 2020GEOM: Energy-annotated molecular conformations for property prediction and molecular generationcited by 243
  3. 2018A General Theory of Equivariant CNNs on Homogeneous Spacescited by 228
  4. 2021Geometric and Physical Quantities Improve E(3) Equivariant Message Passingcited by 228
  5. 2021A Practical Method for Constructing Equivariant Multilayer Perceptrons for Arbitrary Matrix Groupscited by 187
  6. 2018Clebsch-Gordan Nets: a Fully Fourier Space Spherical Convolutional Neural Networkcited by 171
  7. 2021Frame Averaging for Invariant and Equivariant Network Designcited by 168
  8. 2017Equivariance Through Parameter-Sharingcited by 141
  9. 2022Equivariance with Learned Canonicalization Functionscited by 109
  10. 2019Stacked Capsule Autoencoderscited by 82
  11. 2022Benchmarking Robustness of 3D Point Cloud Recognition Against Common Corruptionscited by 80
  12. 2020Meta-Learning Symmetries by Reparameterizationcited by 76
  13. 2021Residual Pathway Priors for Soft Equivariance Constraintscited by 75
  14. 2021Implicit-PDF: Non-Parametric Representation of Probability Distributions on the Rotation Manifoldcited by 65
  15. 2020Attentive Group Equivariant Convolutional Networkscited by 63
  16. 2020Group Equivariant Stand-Alone Self-Attention For Visioncited by 63
  17. 2018Deforming Autoencoders: Unsupervised Disentangling of Shape and Appearancecited by 58
  18. 2020Learning Invariances in Neural Networkscited by 52
  19. 2020On Isometry Robustness of Deep 3D Point Cloud Models under Adversarial Attackscited by 46
  20. 2021Learning Partial Equivariances from Datacited by 45
  21. 2023Generative Adversarial Symmetry Discoverycited by 42
  22. 2023Equivariant Adaptation of Large Pretrained Modelscited by 41
  23. 2022Unsupervised Learning of Group Invariant and Equivariant Representationscited by 39
  24. 2020Probabilistic orientation estimation with matrix Fisher distributionscited by 36
  25. 2023Latent Space Symmetry Discoverycited by 29
  26. 2021Permutation-Invariant Variational Autoencoder for Graph-Level Representation Learningcited by 23
  27. 2023Deep Learning Symmetries and Their Lie Groups, Algebras, and Subalgebras from First Principlescited by 22
  28. 2023A Unified Framework to Enforce, Discover, and Promote Symmetry in Machine Learningcited by 22
  29. 2022Equivariance Discovery by Learned Parameter-Sharingcited by 18
  30. 2022Equivariant Representation Learning via Class-Pose Decompositioncited by 17
  31. 2024Learning Infinitesimal Generators of Continuous Symmetries from Datacited by 17
  32. 2024Symmetry Discovery Beyond Affine Transformationscited by 15
  33. 2021Interpreting Representation Quality of DNNs for 3D Point Cloud Processingcited by 12
  34. 2024A Generative Model of Symmetry Transformationscited by 10
  35. 2019Invariant-equivariant representation learning for multi-class datacited by 6
  36. 2024Learning Symmetries via Weight-Sharing with Doubly Stochastic Tensorscited by 5
  37. 2025Approximating Latent Manifolds in Neural Networks via Vanishing Idealscited by 5
  38. 2023Self-Supervised Detection of Perfect and Partial Input-Dependent Symmetriescited by 4

50 most cited of 54 All 54 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026Neural Field Tokenizations with Hierarchy and Spatial Locality Priorscited by 1