mlsanityData to 14 Oct 2026

On Equivariant Model Selection through the Lens of Uncertainty

arXiv 2506.18629 ↗First posted 23 Jun 2025cs.LG stat.MLcited by 1

Not on the map. Why?

Abstract (arXiv)

Equivariant models leverage prior knowledge on symmetries to improve predictive performance, but misspecified architectural constraints can harm it instead. While work has explored learning or relaxing constraints, selecting among pretrained models with varying symmetry biases remains challenging. We examine this model selection task from an uncertainty-aware perspective, comparing frequentist (via Conformal Prediction), Bayesian (via the marginal likelihood), and calibration-based measures to naive error-based evaluation. We find that uncertainty metrics generally align with predictive performance, but Bayesian model evidence does so inconsistently. We attribute this to a mismatch in Bayesian and geometric notions of model complexity for the employed last-layer Laplace approximation, and discuss possible remedies. Our findings point towards the potential of uncertainty in guiding symmetry-aware model selection.

Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 37 of its 59 references that are in the map. First posted on arXiv 23 Jun 2025.

Builds on

37 of 59 references.

References · 37

  1. 2017On Calibration of Modern Neural Networkscited by 4,960
  2. 2021Geometric Deep Learning: Grids, Groups, Graphs, Geodesics, and Gaugescited by 1,167
  3. 2016Group Equivariant Convolutional Networkscited by 1,165
  4. 2021E(3)-Equivariant Graph Neural Networks for Data-Efficient and Accurate Interatomic Potentialscited by 708
  5. 2021A Survey of Uncertainty in Deep Neural Networkscited by 680
  6. 2016Distribution-Free Predictive Inference For Regressioncited by 668
  7. 2019Measuring Calibration in Deep Learningcited by 446
  8. 2019General E(2)-Equivariant Steerable CNNscited by 446
  9. 2018Accurate Uncertainties for Deep Learning Using Calibrated Regressioncited by 422
  10. 2021Laplace Redux -- Effortless Bayesian Deep Learningcited by 365
  11. 2016Least Ambiguous Set-Valued Classifiers with Bounded Error Levelscited by 272
  12. 2021Geometric and Physical Quantities Improve E(3) Equivariant Message Passingcited by 228
25 more
  1. 2020SalsaNext: Fast, Uncertainty-aware Semantic Segmentation of LiDAR Point Clouds for Autonomous Drivingcited by 226
  2. 2021Geometric Deep Learning on Molecular Representationscited by 113
  3. 2021Scalable Marginal Likelihood Estimation for Model Selection in Deep Learningcited by 106
  4. 2016PAC-Bayesian Theory Meets Bayesian Inferencecited by 88
  5. 2020On the Benefits of Invariance in Neural Networkscited by 79
  6. 2021Residual Pathway Priors for Soft Equivariance Constraintscited by 75
  7. 2021Classifier Calibration: A survey on how to assess and improve predicted class probabilitiescited by 73
  8. 2022Bayesian Model Selection, the Marginal Likelihood, and Generalizationcited by 57
  9. 2018Learning Invariances using the Marginal Likelihoodcited by 51
  10. 2023Approximation-Generalization Trade-offs under (Approximate) Group Equivariancecited by 46
  11. 2022Invariance Learning in Deep Neural Networks with Differentiable Laplace Approximationscited by 42
  12. 2023Fast, Expressive SE(n) Equivariant Networks through Weight-Sharing in Position-Orientation Spacecited by 41
  13. 2020Marginal likelihood computation for model selection and hypothesis testing: an extensive reviewcited by 33
  14. 2023A Large-Scale Study of Probabilistic Calibration in Neural Network Regressioncited by 27
  15. 2023Learning Probabilistic Symmetrization for Architecture Agnostic Equivariancecited by 26
  16. 2023Learning Layer-wise Equivariances Automatically using Gradientscited by 23
  17. 2021Last Layer Marginal Likelihood for Invariance Learningcited by 21
  18. 2020Empirical Frequentist Coverage of Deep Learning Uncertainty Quantification Procedurescited by 19
  19. 2023Regularizing Towards Soft Equivariance Under Mixed Symmetriescited by 16
  20. 2023On genuine invariance learning without weight-tyingcited by 14
  21. 2023Uncertainty Estimation for Molecules: Desiderata and Methodscited by 11
  22. 2020Bayesian Graph Neural Networks for Molecular Property Predictioncited by 8
  23. 2024Learning Symmetries via Weight-Sharing with Doubly Stochastic Tensorscited by 5
  24. 2025CP2: Leveraging Geometry for Conformal Prediction via Canonicalizationcited by 2
  25. 2021From a Point Cloud to a Simulation Model: Bayesian Segmentation and Entropy based Uncertainty Estimation for 3D Modellingcited by 1

Built on by

Cited by year

Citing papers per year: 2025: 1.

1 paper.

  1. 2025On Uncertainty Calibration for Equivariant Functionscited by 1