pith. sign in

arxiv: 2605.01851 · v1 · submitted 2026-05-03 · 📡 eess.SP

Beyond Data-Physics Consistency: A Cross-Correlated Physics-Informed Neural Network for Robust Inverse Scattering

Shilong Sun This is my paper

Pith reviewed 2026-05-08 18:51 UTC · model grok-4.3

classification 📡 eess.SP
keywords inverse scatteringphysics-informed neural networkcross-correlated residualmicrowave imaginghigh-contrast dielectricelectromagnetic inverse problemFourier feature networkGreen's function acceleration
0
0 comments X

The pith

A cross-correlated residual in physics-informed networks couples internal field estimates to external observations, enabling robust reconstruction of high-contrast dielectric targets in inverse scattering.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper proposes a cross-correlated physics-informed neural network to overcome limitations in solving the electromagnetic inverse scattering problem. Standard PINNs minimize separate residuals for the data equation and state equation, which frequently leads to local minima and slow convergence when targets have high dielectric contrast or when observations span multiple frequencies. The new method adds a term that directly correlates the reconstructed permittivity and predicted internal total field with the measured scattered fields, breaking the independence between contrast and field optimization. It also employs a Fourier-feature MLP with weight normalization and accelerates Green's function integration via zero-padded 2D FFT. If the added coupling works as intended, microwave imaging applications could recover complex objects more reliably without relying on careful frequency selection or extensive hyperparameter tuning.

Core claim

The CC-PINN augments the conventional PINN loss with a cross-correlated residual that directly couples the reconstructed dielectric contrast and the predicted internal total electric field to the external scattered-field observations, in addition to the usual data and state residuals. This architecture uses a Fourier-feature multilayer perceptron with weight normalization and a zero-padding 2D-FFT scheme to compute the forward Green's function integrals efficiently. On both synthetic and measured data, the resulting network reconstructs high-contrast dielectric targets with higher fidelity and exhibits convergence robustness superior to standard PINN formulations, whether processing multiple

What carries the argument

The cross-correlated residual term that directly couples the reconstructed dielectric parameters and the predicted internal total field to the external observation field, breaking the decoupling between contrast-source and permittivity optimization.

If this is right

  • High-contrast dielectric targets can be reconstructed with high fidelity from both synthetic and measured scattering data.
  • Convergence remains robust whether the network processes all frequencies simultaneously or uses frequency-hopping strategies.
  • The zero-padding 2D-FFT acceleration lowers the computational cost of repeated Green's function evaluations during training.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same cross-correlation idea could be tested on other wave-equation inverse problems such as acoustic or optical tomography.
  • If the coupling proves stable, practitioners might reduce reliance on multi-frequency data acquisition in field applications.
  • The method invites direct comparison against classical iterative solvers on identical high-contrast benchmarks to quantify speed and accuracy gains.

Load-bearing premise

The added cross-correlated residual will reliably improve the optimization landscape and avoid new instabilities or local minima across varying target contrasts, frequencies, and noise levels without requiring extensive hyperparameter retuning.

What would settle it

A side-by-side optimization run on the same high-contrast synthetic target and noisy multi-frequency measurements where the standard PINN loss reaches a clearly incorrect permittivity distribution while the CC-PINN version recovers the known ground-truth contrast within a small error threshold.

Figures

Figures reproduced from arXiv: 2605.01851 by Shilong Sun.

Figure 6
Figure 6. Figure 6: Final reconstructed images for CC-PINN and classical cost function view at source ↗
Figure 7
Figure 7. Figure 7: Comparison of mean PSNR convergence with standard deviation view at source ↗
Figure 10
Figure 10. Figure 10: Comparison of mean PSNR convergence with standard deviation view at source ↗
Figure 13
Figure 13. Figure 13: Free space scattering measurement facility of the Fresnel data. view at source ↗
Figure 15
Figure 15. Figure 15: Comparison of mean PSNR convergence with standard deviation view at source ↗
Figure 16
Figure 16. Figure 16: Final reconstructed images for CC-PINN and classical cost func view at source ↗
read the original abstract

The electromagnetic inverse scattering problem (ISP), due to its inherent strong nonlinearity and severe ill-posedness, has long been a core challenge in microwave imaging. In recent years, physics-informed neural networks (PINNs) have provided a novel paradigm for solving ISPs by embedding Maxwell's equations into the deep learning optimization process. However, conventional PINN methods rely solely on independent data-equation and state-equation residuals to construct the consistency loss, which easily causes them to fall into local minima and suffer from low computational efficiency when facing high-contrast targets or multi-frequency observation data. To transcend the traditional data-physics consistency framework, this paper proposes a novel cross-correlated physics-informed neural network (CC-PINN). The core innovations of this work include: (1) constructing a Fourier feature MLP network architecture based on weight normalization, which possesses excellent capability for solving inverse scattering problems; (2) introducing a cross-correlated residual term that directly couples the reconstructed dielectric parameters and the predicted internal total field to the external observation field, breaking the decoupling between the contrast source and the permittivity optimization in traditional PINNs and significantly enhancing the robustness of PINNs for ISP; (3) introducing a zero-padding-based 2D-FFT acceleration algorithm, which reduces the computational complexity of the forward Green's function integration. Experimental results on synthetic and measured data demonstrate that CC-PINN can reconstruct high-contrast dielectric targets with high fidelity, and its convergence robustness far exceeds that of PINN algorithms using classical cost functions, regardless of whether simultaneous multi-frequency processing or frequency-hopping strategies are employed.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper proposes a cross-correlated physics-informed neural network (CC-PINN) for electromagnetic inverse scattering problems (ISP). It augments standard PINN formulations with a Fourier-feature MLP using weight normalization, a novel cross-correlated residual term that directly couples reconstructed permittivity, internal total field, and external scattered-field observations to break contrast-source decoupling, and a zero-padding 2D-FFT scheme to accelerate Green's function integration. Experiments on synthetic and measured data are reported to show high-fidelity reconstructions of high-contrast dielectric targets together with markedly improved convergence robustness relative to classical PINN loss functions, for both simultaneous multi-frequency and frequency-hopping data-acquisition strategies.

Significance. If the empirical superiority holds under rigorous controls, the work would meaningfully advance the use of PINNs for strongly nonlinear, ill-posed inverse problems in microwave imaging by mitigating the local-minima trapping that plagues data-physics consistency losses at high contrast. The acceleration technique is a practical engineering contribution. The manuscript does not supply machine-checked proofs or parameter-free derivations, but the introduction of an explicit coupling residual is a clear architectural novelty whose value rests on the strength of the experimental evidence.

major comments (2)
  1. [Methods (cross-correlated residual)] Methods section (cross-correlated residual, Eqs. 8–10): the central robustness claim requires that the new residual improves the loss landscape without introducing new instabilities or necessitating per-instance retuning of its weighting hyperparameter. No supporting analysis—Lipschitz constants, Hessian conditioning, loss-surface visualizations, or systematic sweeps over contrast >5, noise level, or frequency spacing—is provided, leaving open the possibility that observed gains reflect fortunate hyperparameter choices rather than intrinsic superiority.
  2. [Results] Results section: the reported experiments lack quantitative error bars on reconstruction metrics, ablation studies that isolate the contribution of the cross-correlated term, and head-to-head comparisons against standard PINNs under identical network depth, optimizer settings, and data splits. Without these controls the claim that convergence robustness “far exceeds” classical PINNs cannot be evaluated at the level required for the central thesis.
minor comments (2)
  1. [Methods] Notation for the cross-correlated residual is introduced without an explicit statement of how its weighting coefficient is chosen or whether it is held constant across all reported experiments.
  2. [Figures] Figure captions for the reconstruction results should include the exact contrast values, SNR levels, and frequency sets used so that readers can reproduce the claimed operating regime.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments. We agree that strengthening the empirical controls and providing additional analysis of the cross-correlated residual will improve the manuscript and better support our central claims. We address each major comment below and commit to the corresponding revisions.

read point-by-point responses

  1. Referee: [Methods (cross-correlated residual)] Methods section (cross-correlated residual, Eqs. 8–10): the central robustness claim requires that the new residual improves the loss landscape without introducing new instabilities or necessitating per-instance retuning of its weighting hyperparameter. No supporting analysis—Lipschitz constants, Hessian conditioning, loss-surface visualizations, or systematic sweeps over contrast >5, noise level, or frequency spacing—is provided, leaving open the possibility that observed gains reflect fortunate hyperparameter choices rather than intrinsic superiority.

    Authors: We acknowledge that a theoretical characterization of the loss landscape (e.g., Lipschitz bounds or Hessian conditioning) would provide stronger guarantees. The cross-correlated residual (Eqs. 8–10) is explicitly designed to couple the reconstructed permittivity, internal total field, and external scattered-field observations, thereby mitigating the contrast-source decoupling that occurs when data and state residuals are optimized independently. While the original submission did not include formal stability analysis, the reported experiments already span contrasts greater than 5 and multiple noise levels with a fixed weighting hyperparameter schedule. In the revised manuscript we will add (i) loss-surface visualizations for representative high-contrast cases, (ii) systematic sweeps over contrast, additive noise, and frequency spacing, and (iii) a short discussion of observed convergence behavior. These additions will directly address the concern that performance gains may be attributable to per-instance hyperparameter tuning. revision: yes

  2. Referee: [Results] Results section: the reported experiments lack quantitative error bars on reconstruction metrics, ablation studies that isolate the contribution of the cross-correlated term, and head-to-head comparisons against standard PINNs under identical network depth, optimizer settings, and data splits. Without these controls the claim that convergence robustness “far exceeds” classical PINNs cannot be evaluated at the level required for the central thesis.

    Authors: We agree that quantitative error bars, ablation studies, and strictly controlled head-to-head comparisons are necessary to substantiate the robustness claims. The original experiments already demonstrate high-fidelity reconstructions on both synthetic and measured data for high-contrast targets under simultaneous multi-frequency and frequency-hopping acquisition. To meet the referee’s standards, the revised manuscript will include: (i) error bars on all reported metrics (relative permittivity error, scattered-field error, etc.) obtained from multiple independent runs with different random seeds; (ii) ablation experiments that disable only the cross-correlated residual while retaining the Fourier-feature MLP and zero-padding FFT acceleration; and (iii) direct comparisons against standard PINNs using identical network depth, width, optimizer (Adam), learning-rate schedule, and data partitions. These controls will allow an unambiguous assessment of the contribution of the cross-correlated term to convergence robustness. revision: yes

Circularity Check

0 steps flagged

No significant circularity in CC-PINN derivation

full rationale

The paper introduces a cross-correlated residual term, Fourier-feature MLP, and zero-padding FFT acceleration as independent architectural and loss-function innovations that extend standard PINN residuals. These are not defined in terms of each other or prior fitted outputs, nor do any 'predictions' reduce by construction to inputs. Experimental robustness claims rest on new synthetic/measured data trials rather than self-citation chains or renamed empirical patterns. The derivation chain remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The method rests on Maxwell's equations as the governing physics and standard neural network optimization assumptions; the cross-correlated term is the primary new element introduced without external validation.

axioms (1)
  • domain assumption Maxwell's equations accurately describe the electromagnetic fields in the scattering scenario
    Invoked as the physics residual embedded in the network loss, standard for PINN approaches to ISP.
invented entities (1)
  • cross-correlated residual term no independent evidence
    purpose: To directly couple reconstructed dielectric parameters and predicted internal total field to the external observation field
    New loss component introduced to break the decoupling present in traditional independent data and state equation residuals.

pith-pipeline@v0.9.0 · 5581 in / 1301 out tokens · 39963 ms · 2026-05-08T18:51:20.067886+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

41 extracted references · 41 canonical work pages

  1. [1]

    Inversion of seismic reflection data in the acoustic approximation,

    A. Tarantola, “Inversion of seismic reflection data in the acoustic approximation,”Geophysics, vol. 49, no. 8, pp. 1259–1266, 1984

    work page 1984

  2. [2]

    Seismic inversion,

    G. T. Schuster, “Seismic inversion,”Society of Exploration Geophysi- cists, 2017

    work page 2017

  3. [3]

    Distinguishability of conductivities by electric current computed tomography,

    D. Isaacson, “Distinguishability of conductivities by electric current computed tomography,”IEEE Transactions on Medical Imaging, vol. 5, no. 2, pp. 91–95, 1986

    work page 1986

  4. [4]

    Nonlinear inverse problems in imaging,

    J. K. Seo and E. J. Woo, “Nonlinear inverse problems in imaging,” Annual Review of Biomedical Engineering, vol. 5, pp. 413–449, 2003

    work page 2003

  5. [5]

    Electrical impedance tomography,

    M. Cheney, D. Isaacson, and J. C. Newell, “Electrical impedance tomography,”SIAM Review, vol. 41, no. 1, pp. 85–101, 2000

    work page 2000

  6. [6]

    Colton and R

    D. Colton and R. Kress,Inverse acoustic and electromagnetic scattering theory, vol. 93. Springer Science & Business Media, 2012

    work page 2012

  7. [7]

    Imaging with diffraction tomography,

    M. Slaney and A. C. Kak, “Imaging with diffraction tomography,” Tech. Rep. TR-EE 85-5, Department of Electrical and Computer Engineering, Purdue University, West Lafayette, Indiana 47907, February 1985

    work page 1985

  8. [8]

    Novel diffraction tomographic algorithm for imaging two-dimensional targets buried under a lossy earth,

    T. J. Cui and W. C. Chew, “Novel diffraction tomographic algorithm for imaging two-dimensional targets buried under a lossy earth,”IEEE Transactions on Geoscience and Remote Sensing, vol. 38, no. 4, pp. 2033–2041, 2000

    work page 2033

  9. [9]

    Quantitative diffraction tomography for weak scatterers based on aliasing modification of the multifrequency spatial spectrum,

    M. Wang, S. Sun, D. Dai, Y . Su, and M. Wu, “Quantitative diffraction tomography for weak scatterers based on aliasing modification of the multifrequency spatial spectrum,”IEEE Transactions on Geoscience and Remote Sensing, vol. 61, pp. 1–14, 2023

    work page 2023

  10. [10]

    A contrast source inversion method,

    P. M. Van Den Berg and R. E. Kleinman, “A contrast source inversion method,”Inverse Problems, vol. 13, no. 6, pp. 1607–1620, 1997

    work page 1997

  11. [11]

    A subspace-based optimization method for solving inverse scattering problems,

    J. Li, A. Abubakar, and P. M. van den Berg, “A subspace-based optimization method for solving inverse scattering problems,”IEEE Transactions on Geoscience and Remote Sensing, vol. 48, no. 1, pp. 42– 49, 2010

    work page 2010

  12. [12]

    Subspace-based optimization method for solving inverse- scattering problems,

    X. Chen, “Subspace-based optimization method for solving inverse- scattering problems,”IEEE Transactions on Geoscience and Remote Sensing, vol. 48, no. 1, pp. 42–49, 2010

    work page 2010

  13. [13]

    Cross-correlated subspace-based optimization method for solving electromagnetic inverse scattering problems,

    M. Wang, S. Sun, D. Dai, Y . Zhang, and Y . Su, “Cross-correlated subspace-based optimization method for solving electromagnetic inverse scattering problems,”IEEE Transactions on Antennas and Propagation, vol. 72, no. 11, pp. 8575–8589, 2024

    work page 2024

  14. [14]

    Cross-correlated contrast source inversion,

    S. Sun, B. J. Kooij, T. Jin, and A. G. Yarovoy, “Cross-correlated contrast source inversion,”IEEE Transactions on Antennas and Propagation, vol. 65, no. 5, pp. 2592–2603, 2017

    work page 2017

  15. [15]

    Inversion of multifrequency data with the cross-correlated contrast source inversion method,

    S. Sun, B.-J. Kooij, and A. G. Yarovoy, “Inversion of multifrequency data with the cross-correlated contrast source inversion method,”Radio Science, vol. 53, no. 6, pp. 710–723, 2018

    work page 2018

  16. [16]

    Frequency-binned cumulative hopping framework incorporating wavelength-dependent weighting strategies for inverse scattering of high-contrast objects,

    S. Sun, “Frequency-binned cumulative hopping framework incorporating wavelength-dependent weighting strategies for inverse scattering of high-contrast objects,”IEEE Transactions on Antennas and Propagation, vol. 73, no. 10, pp. 8048–8062, 2025

    work page 2025

  17. [17]

    Reconstruction of two-dimensional permittivity distribution using the distorted Born iterative method,

    W. C. Chew and Y .-M. Wang, “Reconstruction of two-dimensional permittivity distribution using the distorted Born iterative method,”IEEE Transactions on Medical Imaging, vol. 9, no. 2, pp. 218–225, 1990

    work page 1990

  18. [18]

    Multiple buried target reconstruction using a multiscale hybrid of diffraction tomography and CMA-ES optimization,

    M. Hajebi and A. Hoorfar, “Multiple buried target reconstruction using a multiscale hybrid of diffraction tomography and CMA-ES optimization,” IEEE Transactions on Geoscience and Remote Sensing, vol. 60, pp. 1– 13, 2022

    work page 2022

  19. [19]

    Subsurface inverse profiling and imaging us- ing stochastic optimization techniques,

    M. Hajebi and A. Hoorfar, “Subsurface inverse profiling and imaging us- ing stochastic optimization techniques,” inSignal and Image Processing for Remote Sensing, pp. 45–75, CRC Press, 2024

    work page 2024

  20. [20]

    Stochastic optimization methods applied to microwave imaging: A review,

    M. Pastorino, “Stochastic optimization methods applied to microwave imaging: A review,”IEEE Transactions on Antennas and Propagation, vol. 55, no. 3, pp. 538–548, 2007

    work page 2007

  21. [21]

    Electromagnetic detection of dielectric cylin- ders by a neural network approach,

    S. Caorsi and P. Gamba, “Electromagnetic detection of dielectric cylin- ders by a neural network approach,”IEEE Transactions on Geoscience and Remote Sensing, vol. 37, no. 2, pp. 820–827, 1999

    work page 1999

  22. [22]

    Efficient and accurate inversion of multiple scattering with deep learning,

    Y . Sun, Z. Xia, and U. S. Kamilov, “Efficient and accurate inversion of multiple scattering with deep learning,”Opt. Express, vol. 26, pp. 14678–14688, May 2018

    work page 2018

  23. [23]

    Deepnis: Deep neural network for nonlinear electromagnetic inverse scattering,

    L. Li, L. G. Wang, F. L. Teixeira, C. Liu, A. Nehorai, and T. J. Cui, “Deepnis: Deep neural network for nonlinear electromagnetic inverse scattering,”IEEE Transactions on Antennas and Propagation, vol. 67, no. 3, pp. 1819–1825, 2019

    work page 2019

  24. [24]

    Solving electromagnetic inverse scattering problems in inhomogeneous media by deep convolu- tional encoder–decoder structure,

    H. H. Zhang, H. M. Yao, L. Jiang, and M. Ng, “Solving electromagnetic inverse scattering problems in inhomogeneous media by deep convolu- tional encoder–decoder structure,”IEEE Transactions on Antennas and Propagation, vol. 71, no. 3, pp. 2867–2872, 2023

    work page 2023

  25. [25]

    Learning-based profiling of buried elliptical-cylindrical objects,

    Z. Dastfal, M. Hajebi, M. Sharifzadeh, and A. Hoorfar, “Learning-based profiling of buried elliptical-cylindrical objects,”IEEE Geoscience and Remote Sensing Letters, vol. 22, pp. 1–5, 2025

    work page 2025

  26. [26]

    PUP-Net: A twofold physical model embedded 3-D U-Net with polarization fusion for solving inverse scattering problems with a sparse planar array,

    M. Wang, S. Sun, Y . Zhang, D. Dai, H. Wu, and Y . Su, “PUP-Net: A twofold physical model embedded 3-D U-Net with polarization fusion for solving inverse scattering problems with a sparse planar array,”IEEE Transactions on Microwave Theory and Techniques, vol. 73, no. 4, pp. 2123–2136, 2025

    work page 2025

  27. [27]

    A non- linear model compression scheme based on variational autoencoder for microwave data inversion,

    R. Guo, Z. Lin, M. Li, F. Yang, S. Xu, and A. Abubakar, “A non- linear model compression scheme based on variational autoencoder for microwave data inversion,”IEEE Transactions on Antennas and Propagation, vol. 70, no. 11, pp. 11059–11069, 2022

    work page 2022

  28. [28]

    Probabilistic deep learning solutions to electromagnetic inverse scattering problems using conditional renormalization group flow,

    H. Zhang, Y . Chen, T. J. Cui, F. L. Teixeira, and L. Li, “Probabilistic deep learning solutions to electromagnetic inverse scattering problems using conditional renormalization group flow,”IEEE Transactions on Microwave Theory and Techniques, vol. 70, no. 11, pp. 4955–4965, 2022

    work page 2022

  29. [29]

    Inhomogeneous media inverse scattering problem assisted by swin transformer network,

    N. Du, J. Wang, R. Song, K. Xu, S. Sun, and X. Ye, “Inhomogeneous media inverse scattering problem assisted by swin transformer network,” IEEE Transactions on Microwave Theory and Techniques, vol. 72, no. 12, pp. 6809–6820, 2024

    work page 2024

  30. [30]

    Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations,

    M. Raissi, P. Perdikaris, and G. Karniadakis, “Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations,”Journal of Computational Physics, vol. 378, pp. 686–707, 2019

    work page 2019

  31. [31]

    Physics-informed neural networks for inverse problems in nano-optics and metamaterials.,

    Y . Chen, L. Lu, G. E. Karniadakis, and L. D. Negro, “Physics-informed neural networks for inverse problems in nano-optics and metamaterials.,” Optics express, vol. 28 8, pp. 11618–11633, 2019

    work page 2019

  32. [32]

    Weight normalization: a simple reparameterization to accelerate training of deep neural networks,

    T. Salimans and D. P. Kingma, “Weight normalization: a simple reparameterization to accelerate training of deep neural networks,” in Proceedings of the 30th International Conference on Neural Information Processing Systems, NIPS’16, (Red Hook, NY , USA), p. 901–909, Curran Associates Inc., 2016

    work page 2016

  33. [33]

    Random features for large-scale kernel machines,

    A. Rahimi and B. Recht, “Random features for large-scale kernel machines,” inAdvances in Neural Information Processing Systems 20, Proceedings of the Twenty-First Annual Conference on Neural Information Processing Systems, Vancouver, British Columbia, Canada, December 3-6, 2007(J. C. Platt, D. Koller, Y . Singer, and S. T. Roweis, eds.), pp. 1177–1184, Cu...

    work page 2007

  34. [34]

    Fourier features let networks learn high frequency functions in low dimensional domains,

    M. Tancik, P. Srinivasan, B. Mildenhall, S. Fridovich-Keil, N. Raghavan, U. Singhal, R. Ramamoorthi, J. Barron, and R. Ng, “Fourier features let networks learn high frequency functions in low dimensional domains,” inAdvances in Neural Information Processing Systems(H. Larochelle, M. Ranzato, R. Hadsell, M. Balcan, and H. Lin, eds.), vol. 33, pp. 7537– 754...

    work page 2020

  35. [35]

    A fast algorithm of cross-correlated contrast source inversion in homogeneous background media,

    S. Sun, D. Dai, and X. Wang, “A fast algorithm of cross-correlated contrast source inversion in homogeneous background media,”IEEE Transactions on Antennas and Propagation, vol. 71, no. 5, pp. 4380– 4393, 2023

    work page 2023

  36. [36]

    Shin,3D finite-difference frequency-domain method for plasmonics and nanophotonics

    W. Shin,3D finite-difference frequency-domain method for plasmonics and nanophotonics. PhD thesis, Stanford University, 2013

    work page 2013

  37. [37]

    Using multiple frequency information in the iterative solution of a two-dimensional nonlinear inverse problem,

    K. Belkebir and A. Tijhuis, “Using multiple frequency information in the iterative solution of a two-dimensional nonlinear inverse problem,” in Proceedings Progress in Electromagnetics Research Symposium, PIERS 1996, 8 July 1996, Innsbruck, Germany, p. 353, University of Innsbruck, 1996

    work page 1996

  38. [38]

    Reconstruction of a two- dimensional binary obstacle by controlled evolution of a level-set,

    A. Litman, D. Lesselier, and F. Santosa, “Reconstruction of a two- dimensional binary obstacle by controlled evolution of a level-set,” Inverse Problems, vol. 14, no. 3, pp. 685–706, 1998

    work page 1998

  39. [39]

    Contrast source inversion method: state of art,

    P. M. van den Berg and A. Abubakar, “Contrast source inversion method: state of art,”Journal of Electromagnetic Waves and Applications, vol. 15, no. 11, pp. 1503–1505, 2001

    work page 2001

  40. [40]

    Multiplicative regularization for contrast profile inversion,

    P. M. van den Berg, A. Abubakar, and J. T. Fokkema, “Multiplicative regularization for contrast profile inversion,”Radio Science, vol. 38, no. 2, 2003

    work page 2003

  41. [41]

    Free space experimental scattering database continuation: experimental set-up and measurement precision,

    J.-M. Geffrin, P. Sabouroux, and C. Eyraud, “Free space experimental scattering database continuation: experimental set-up and measurement precision,”Inverse Problems, vol. 21, no. 6, pp. S117–S130, 2005

    work page 2005