mlsanityData to 7 Oct 2026

Discretization Invariant Networks for Learning Maps between Neural Fields

arXiv 2206.01178 ↗First posted 2 Jun 2022cs.LG cs.CV cs.NEcited by 0

Not on the map. Why?

Abstract (arXiv)

With the emergence of powerful representations of continuous data in the form of neural fields, there is a need for discretization invariant learning: an approach for learning maps between functions on continuous domains without being sensitive to how the function is sampled. We present a new framework for understanding and designing discretization invariant neural networks (DI-Nets), which generalizes many discrete networks such as convolutional neural networks as well as continuous networks such as neural operators. Our analysis establishes upper bounds on the deviation in model outputs under different finite discretizations, and highlights the central role of point set discrepancy in characterizing such bounds. This insight leads to the design of a family of neural networks driven by numerical integration via quasi-Monte Carlo sampling with discretizations of low discrepancy. We prove by construction that DI-Nets universally approximate a large class of maps between integrable function spaces, and show that discretization invariance also describes backpropagation through such models. Applied to neural fields, convolutional DI-Nets can learn to classify and segment visual data under various discretizations, and sometimes generalize to new types of discretizations at test time. Code: https://github.com/clintonjwang/DI-net.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 43 of its 73 references that are in the map. First posted on arXiv 2 Jun 2022.

Builds on

43 of 73 references.

References · 43

  1. 2019EfficientNet: Rethinking Model Scaling for Convolutional Neural Networkscited by 6,415
  2. 2016The Cityscapes Dataset for Semantic Urban Scene Understandingcited by 4,427
  3. 2017PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Spacecited by 4,384
  4. 2022Instant Neural Graphics Primitives with a Multiresolution Hash Encodingcited by 3,698
  5. 2022A ConvNet for the 2020scited by 3,545
  6. 2020Fourier Neural Operator for Parametric Partial Differential Equationscited by 2,806
  7. 2019DeepSDF: Learning Continuous Signed Distance Functions for Shape Representationcited by 2,682
  8. 2020Implicit Neural Representations with Periodic Activation Functionscited by 2,214
  9. 2020Fourier Features Let Networks Learn High Frequency Functions in Low Dimensional Domainscited by 2,170
  10. 2018Occupancy Networks: Learning 3D Reconstruction in Function Spacecited by 2,037
  11. 2019DeepONet: Learning nonlinear operators for identifying differential equations based on the universal approximation theorem of operatorscited by 1,965
  12. 2017Decoupled Weight Decay Regularizationcited by 1,722
31 more
  1. 2021Plenoxels: Radiance Fields without Neural Networkscited by 1,607
  2. 2017Deep Setscited by 1,394
  3. 2016Group Equivariant Convolutional Networkscited by 1,154
  4. 2021Direct Voxel Grid Optimization: Super-fast Convergence for Radiance Fields Reconstructioncited by 961
  5. 2021PlenOctrees for Real-time Rendering of Neural Radiance Fieldscited by 904
  6. 2019Scene Representation Networks: Continuous 3D-Structure-Aware Neural Scene Representationscited by 864
  7. 2020Neural Operator: Graph Kernel Network for Partial Differential Equationscited by 722
  8. 2020pi-GAN: Periodic Implicit Generative Adversarial Networks for 3D-Aware Image Synthesiscited by 708
  9. 2019Differentiable Volumetric Rendering: Learning Implicit 3D Representations without 3D Supervisioncited by 699
  10. 2021Neural Operator: Learning Maps Between Function Spacescited by 413
  11. 2021Light Field Networks: Neural Scene Representations with Single-Evaluation Renderingcited by 330
  12. 2017The Expressive Power of Neural Networks: A View from the Widthcited by 318
  13. 2016Steerable CNNscited by 298
  14. 2020Model Reduction and Neural Networks for Parametric PDEscited by 257
  15. 2022Panoptic Neural Fields: A Semantic Object-Aware Neural Scene Representationcited by 243
  16. 2020Learned Initializations for Optimizing Coordinate-Based Neural Representationscited by 240
  17. 2020MetaSDF: Meta-learning Signed Distance Functionscited by 214
  18. 2020DiffusionNet: Discretization Agnostic Learning on Surfacescited by 163
  19. 2021Deep Parametric Continuous Convolutional Neural Networkscited by 161
  20. 2017SplineCNN: Fast Geometric Deep Learning with Continuous B-Spline Kernelscited by 160
  21. 2021Modulated Periodic Activations for Generalizable Local Functional Representationscited by 140
  22. 2018Universal approximations of invariant maps by neural networkscited by 127
  23. 2021NeSF: Neural Semantic Fields for Generalizable Semantic Segmentation of 3D Scenescited by 119
  24. 2021CKConv: Continuous Kernel Convolution For Sequential Datacited by 117
  25. 2021Generative Models as Distributions of Functionscited by 89
  26. 2020On the Benefits of Invariance in Neural Networkscited by 78
  27. 2021ObjectFolder: A Dataset of Objects with Implicit Visual, Auditory, and Tactile Representationscited by 67
  28. 2021FlexConv: Continuous Kernel Convolutions with Differentiable Kernel Sizescited by 66
  29. 2022AutoRF: Learning 3D Object Radiance Fields from Single View Observationscited by 59
  30. 2020A Functional Perspective on Learning Symmetric Functions with Neural Networkscited by 16
  31. 2019Convolutional Neural Networks on non-uniform geometrical signals using Euclidean spectral transformationcited by 3

Built on by

Not cited yet.