mlsanityData to 7 Oct 2026

What Affects Learned Equivariance in Deep Image Recognition Models?

arXiv 2304.02628 ↗First posted 5 Apr 2023cs.CV cs.AIcited by 14

Not on the map. Why?

Abstract (arXiv)

Equivariance w.r.t. geometric transformations in neural networks improves data efficiency, parameter efficiency and robustness to out-of-domain perspective shifts. When equivariance is not designed into a neural network, the network can still learn equivariant functions from the data. We quantify this learned equivariance, by proposing an improved measure for equivariance. We find evidence for a correlation between learned translation equivariance and validation accuracy on ImageNet. We therefore investigate what can increase the learned equivariance in neural networks, and find that data augmentation, reduced model capacity and inductive bias in the form of convolutions induce higher learned equivariance in neural networks.

Key facts. Cited by 14 papers in the map (data through 7 Oct 2026). Builds on 32 of its 48 references that are in the map. First posted on arXiv 5 Apr 2023.

Builds on

32 of 48 references.

References · 32

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,160
  3. 2019EfficientNet: Rethinking Model Scaling for Convolutional Neural Networkscited by 6,415
  4. 2015Rethinking the Inception Architecture for Computer Visioncited by 5,451
  5. 2016Wide Residual Networkscited by 3,057
  6. 2019CutMix: Regularization Strategy to Train Strong Classifiers with Localizable Featurescited by 2,875
  7. 2016Aggregated Residual Transformations for Deep Neural Networkscited by 2,540
  8. 2019RandAugment: Practical automated data augmentation with a reduced search spacecited by 2,261
  9. 2019Similarity of Neural Network Representations Revisitedcited by 1,695
  10. 2021MLP-Mixer: An all-MLP Architecture for Visioncited by 1,441
  11. 2016Group Equivariant Convolutional Networkscited by 1,154
  12. 2021ConViT: Improving Vision Transformers with Soft Convolutional Inductive Biasescited by 484
20 more
  1. 2019General E(2)-Equivariant Steerable CNNscited by 441
  2. 2019Making Convolutional Networks Shift-Invariant Againcited by 390
  3. 2019Gauge Equivariant Convolutional Networks and the Icosahedral CNNcited by 262
  4. 2014Understanding image representations by measuring their equivariance and equivalencecited by 240
  5. 2021ViTAE: Vision Transformer Advanced by Exploring Intrinsic Inductive Biascited by 216
  6. 2017Learning Steerable Filters for Rotation Equivariant CNNscited by 193
  7. 2020On Translation Invariance in CNNs: Convolutional Layers can Exploit Absolute Spatial Locationcited by 141
  8. 2016Exploiting Cyclic Symmetry in Convolutional Neural Networkscited by 91
  9. 20183D G-CNNs for Pulmonary Nodule Detectioncited by 36
  10. 2017Quantifying Translation-Invariance in Convolutional Neural Networkscited by 30
  11. 2022LieGG: Studying Learned Lie Group Generatorscited by 28
  12. 2021Grounding inductive biases in natural images:invariance stems from variations in datacited by 25
  13. 2022Scale Equivariant U-Netcited by 13
  14. 2021Global Pooling, More than Meets the Eye: Position Information is Encoded Channel-Wise in CNNscited by 12
  15. 2022Do Deep Networks Transfer Invariances Across Classes?cited by 12
  16. 2018Pooling is neither necessary nor sufficient for appropriate deformation stability in CNNscited by 8
  17. 2020Learning Translation Invariance in CNNscited by 8
  18. 2021Exploiting Learned Symmetries in Group Equivariant Convolutionscited by 5
  19. 2021Tracking translation invariance in CNNscited by 3
  20. 2022Just a Matter of Scale? Reevaluating Scale Equivariance in Convolutional Neural Networkscited by 2

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 3; 2024: 4; 2025: 2; 2026: 5.

14 papers.

Sort:

By area:

  1. 2023Steerers: A framework for rotation equivariant keypoint descriptorscited by 21
  2. 2023Investigating how ReLU-networks encode symmetriescited by 8
  3. 2025Flopping for FLOPs: Leveraging equivariance for computational efficiencycited by 8
  4. 2023Color Equivariant Convolutional Networkscited by 7
  5. 2024Affine steerers for structured keypoint descriptioncited by 6
  6. 2024Don't Think It Twice: Exploit Shift Invariance for Efficient Online Streaming Inference of CNNscited by 1
  7. 2025EquiReg: Equivariance Regularized Diffusion for Inverse Problemscited by 1
  8. 2026GEPC: Group-Equivariant Posterior Consistency for Out-of-Distribution Detection in Diffusion Modelscited by 1
  9. 2026Stimulus symmetries can confound representational similarity analysescited by 1
  10. 2024Adapting Pretrained ViTs with Convolution Injector for Visuo-Motor Controlcited by 0
  11. 2024MARs: Multi-view Attention Regularizations for Patch-based Feature Recognition of Space Terraincited by 0
  12. 2026SEIS: Subspace-based Equivariance and Invariance Scores for Neural Representationscited by 0
2 more
  1. 2026Scale-invariant Gaussian derivative residual networkscited by 0
  2. 2026Transformation Laws in Neural Representations: Structure, Realisability, and Constructioncited by 0