mlsanityData to 14 Oct 2026

Self-composing neural operators for high-frequency and multiscale PDE surrogates

arXiv 2508.20650 ↗First posted 28 Aug 2025cs.LGcited by 3

Not on the map. Why?

Abstract (arXiv)

Addressing the computational challenges of high-frequency and multiscale partial differential equations (PDEs), this work introduces a self-composing neural operator (SC-NO) framework. Inspired by classical fixed-point iterative solvers (e.g., multigrid, domain decomposition), the proposed architecture constructs a deep operator by repeatedly applying a single, parameter-efficient backbone block. This design mimics the update step of a numerical solver, allowing the model to progressively resolve complex solution features without increasing the parameter count. For practical training, we develop an adaptive ``Train-and-Unroll'' strategy that grows the composition depth during training, acting as a curriculum from shallow to deep self-composed models. We demonstrate the efficacy of this framework on the Helmholtz equation for ultrasound computed tomography (USCT), a problem characterized by high-frequency wave propagation in highly heterogeneous media. By instantiating the backbone with a multigrid-inspired architecture, the SC-NO effectively mitigates the spectral bias often observed in standard operator learning baselines. Numerical experiments show that our method reduces the prediction error significantly compared to Fourier Neural Operators (FNO) and their variants in the 300--500 kHz regime. Furthermore, we provide theoretical analysis linking the self-composition depth to approximation accuracy.

Key facts. Cited by 3 papers in the map (data through 14 Oct 2026). Builds on 37 of its 67 references that are in the map. First posted on arXiv 28 Aug 2025.

Builds on

37 of 67 references.

References · 37

  1. 2017Attention Is All You Needcited by 60,361
  2. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  3. 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,713
  4. 2018Neural Ordinary Differential Equationscited by 3,779
  5. 2019DeepONet: Learning nonlinear operators for identifying differential equations based on the universal approximation theorem of operatorscited by 2,018
  6. 2016Identity Mappings in Deep Residual Networkscited by 1,965
  7. 2019Deep Equilibrium Modelscited by 615
  8. 2022Fourier Neural Operator with Learned Deformations for PDEs on General Geometriescited by 427
  9. 2021Choose a Transformer: Fourier or Galerkincited by 343
  10. 2023GNOT: A General Neural Operator Transformer for Operator Learningcited by 336
  11. 2021Multiwavelet-based Operator Learning for Differential Equationscited by 234
  12. 2023Convolutional Neural Operators for robust and accurate learning of PDEscited by 214
25 more
  1. 2017Beyond Finite Layer Neural Networks: Bridging Deep Architectures and Numerical Differential Equationscited by 203
  2. 2022MIONet: Learning multiple-input operators via tensor productcited by 181
  3. 2020PhyGeoNet: Physics-Informed Geometry-Adaptive Convolutional Neural Networks for Solving Parameterized Steady-State PDEs on Irregular Domaincited by 174
  4. 2025Transolver++: An Accurate Neural Solver for PDEs on Million-Scale Geometriescited by 97
  5. 2023Solving High-Dimensional PDEs with Latent Spectral Modelscited by 89
  6. 2021JFB: Jacobian-Free Backpropagation for Implicit Networkscited by 86
  7. 2021On Training Implicit Modelscited by 75
  8. 2022Blending Neural Operators and Relaxation Methods in PDE Numerical Solverscited by 54
  9. 2022Multi-resolution partial differential equations preserved learning framework for spatiotemporal dynamicscited by 44
  10. 2022Mitigating spectral bias for the multiscale operator learningcited by 44
  11. 2019MgNet: A Unified Framework of Multigrid and Convolutional Neural Networkcited by 41
  12. 2023MgNO: Efficient Parameterization of Linear Operators via Multigridcited by 40
  13. 2018SwitchNet: a neural network model for forward and inverse scattering problemscited by 35
  14. 2023Out-of-distributional risk bounds for neural operators with applications to the Helmholtz equationcited by 32
  15. 2022Multigrid-augmented deep learning preconditioners for the Helmholtz equationcited by 25
  16. 2022A Physics-Guided Neural Operator Learning Approach to Model Biological Tissues from Digital Image Correlation Measurementscited by 20
  17. 2023Deep Equilibrium Based Neural Operators for Steady-State PDEscited by 19
  18. 2023IB-UQ: Information bottleneck based uncertainty quantification for neural function regression and neural operator learningcited by 18
  19. 2025Geometric Operator Learning with Optimal Transportcited by 16
  20. 2024A hybrid iterative method based on MIONet for PDEs: Theory and numerical examplescited by 13
  21. 2023Multigrid-Augmented Deep Learning Preconditioners for the Helmholtz Equation using Compact Implicit Layerscited by 11
  22. 2021An Interpretive Constrained Linear Model for ResNet and MgNetcited by 6
  23. 2023On Enhancing Expressive Power via Compositions of Single Fixed-Size ReLU Networkcited by 6
  24. 2023An Enhanced V-cycle MgNet Model for Operator Learning in Numerical Partial Differential Equationscited by 4
  25. 2023FV-MgNet: Fully Connected V-cycle MgNet for Interpretable Time Series Forecastingcited by 2

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 1; 2026: 2.

3 papers.

  1. 2025Mechanistic Interpretability with Sparse Autoencoder Neural Operatorscited by 3
  2. 2026Neural Flow Operators can Approximate any Operator: Abstract Frameworks and Universal Approximationscited by 0
  3. 2026Starter-Iterator Neural Operator: A Unified Architecture for High-Fidelity Forward and Inverse PDE Problemscited by 0