betaFeedback

WorksPhysics and Astronomy

NAP-Wave: A Neural Adaptive Plane-Wave Architecture for High-Frequency Helmholtz Problems

Abstract

Training neural networks to simulate highly oscillatory wavefields presents optimization challenges, often hindered by spectral bias and poorly conditioned physics-based loss landscapes. To address these challenges and enhance convergence, we propose NAP-Wave, a joint neural architecture and training strategy that expands the target scattered wavefield in an adaptive neural plane-wave basis. The basis functions are extracted from all hidden layers and modulated by trainable plane-wave carriers, embedding the high-frequency oscillatory physics directly into the network's trial space. NAP-Wave is trained on a physics-informed loss built from the Lippmann-Schwinger integral equation, so no differentiation of the network with respect to its input is required. We develop a decoupled optimization strategy in which Adam updates the parameters of the neural plane-wave basis, while the linear combination coefficients are computed exactly by a least-squares (LS) solver. We couple this with an adaptive-depth training strategy in which the network starts shallow and grows progressively layer-by-layer. At each growth step, we avoid the loss spikes typical of dynamic network growth using the LS-computed coefficients, the warm-started parameters, and the preservation of the first- and second-order optimizer moments of the previous stage. Ablation studies confirm that the plane-wave basis, the exact coefficient solve, and the progressive schedule independently accelerate convergence and improve accuracy. Validated on exact-solution benchmarks, NAP-Wave achieves relative L2 errors of order 1%. On highly heterogeneous geophysical benchmarks, it achieves relative errors of 5%, 9%, and 10%, significantly outperforming published PINN methods in both accuracy and convergence time. To ensure reproducibility, the code is available at https://github.com/Hamza-Kamil/NAP-WAVE.

Introduction

Background and motivation

Neural PDE solvers have emerged as promising mesh-free algorithms for scientific simulations (e.g., [1, 3, 29, 24]). However, learning highly oscillatory, high-frequency functions remains a profound challenge for standard neural architectures. This difficulty is mainly driven by spectral bias, a phenomenon where neural networks inherently prioritize learning low-frequency components and struggle to fit high-frequency details [27]. Furthermore, minimizing high-frequency PDE residuals often requires computing second-order spatial derivatives, which can result in highly non-convex, poorly conditioned loss landscapes that severely impede optimizer convergence.

A particularly challenging setting for testing these neural solvers is time-harmonic wave propagation, governed by the Helmholtz equation. As the frequency and medium contrast increase, the exact solutions become highly oscillatory and multiscale. While classical discretizations require prohibitively large, indefinite linear systems to resolve these fine wavelengths, standard deep neural networks equally fail due to spectral bias and the difficulty of resolving sharp localized features in highly heterogeneous velocity models, such as those commonly found in geophysics and medical imaging [6, 14].

The Lippmann-Schwinger integral formulation eliminates second-order derivatives that appear in the strong form of the Helmholtz equation. As a result, the training loss does not require differentiation of the network with respect to its input, and the associated integral operator is better conditioned than the differential Helmholtz operator. Embedding the outgoing radiation condition into the background Green’s function turns the scattering problem into a Fredholm integral equation of the second kind. The differential Helmholtz operator is unbounded and ill-conditioned, whereas the resulting integral operator is a bounded identity-plus-compact operator. This structural preconditioning gives a smoother loss landscape and helps stabilize training. The action of the Green operator on the trial space is evaluated on a uniform grid via FFT, exploiting the Toeplitz structure of the kernel. This evaluation is far faster than direct evaluation of the Helmholtz operator, which requires differentiating the network at every collocation point.

Abedi et al. [2] proposed the Lippmann-Schwinger integral equation as a training loss for neural solvers of scattering problems, combining it with an auxiliary strong-form residual as a regularizer. We adopt the same Green-integral loss and instead improve the architecture and optimization strategy: for the same loss, our design converges faster and reaches lower error in strongly heterogeneous media. The NN architecture of the trial space is designed so that the wave structure is represented directly by the network, and it is paired with an optimization strategy built specifically to make training fast and stable.

In this study, we introduce NAP-Wave, a data-free neural solver built on three design principles.

  1. The wave structure belongs in an adaptive plane-wave trial space. A generic network must discover, from a smooth coordinate input, a field oscillating on the scale of a wavelength. Trefftz and plane-wave methods avoid this by building basis functions that satisfy the governing equation exactly [17, 8]. We adopt the same principle in a neural setting. Each basis function carries a plane-wave that solves the homogeneous Helmholtz equation exactly, and the network represents the varying amplitude that modulates it.

  2. The linear coefficients are solved exactly, while the basis adapts. The Lippmann-Schwinger operator is linear, so the coefficients combining the basis functions admit a closed-form solution for any fixed basis. We solve for them by least squares at every step. Adam updates only the nonlinear parameters that shape the basis functions themselves. This is the classical variable projection principle [13], applied here with every neuron of every hidden layer treated as an individual basis function.

  3. The trial space grows with training, rather than staying fixed. A network initialized at full depth must resolve a high-wavenumber field from a random basis. We instead train a sequence of progressively deeper sub-networks. Each new layer adds a block of basis functions and preserves the first- and second-order moments of the optimizer from the converged parameters of the previous stage, so that the trial space is enriched only once the coarser one has been exploited. Networks are known to learn low-frequency components first [27]; growing depth progressively lets each new layer target the higher-frequency content the shallower network could not yet represent. In practice, we find that this progressive schedule acts as a warm start: once a shallow sub-network has converged, its parameters already place the next, deeper sub-network in a good basis, and adding further training iterations to the shallow network alone yields no additional gain.

Related work

Integral-equation solvers for scattering. The Lippmann-Schwinger equation is a classical tool in scattering theory. It appears across many fields, including quantum mechanics [28], electromagnetics [9], acoustics [22], and seismic wave propagation [18], and we adopt it here for frequency-domain seismic scattering. Its mathematical foundations are well established, including compactness of the volume potential, Fredholm invertibility, and radiation conditions [20]. Traditional numerical solvers discretize the integral operator and assemble the resulting dense system. This system is prohibitively expensive to invert directly. Practitioners therefore use iterative solvers instead, in which each matrix-vector product is evaluated efficiently via the FFT. This approach exploits the Toeplitz structure of the Green’s kernel on uniform grids. The Born series [23] is a classical instance of this iterative approach. It converges for weak scattering contrasts, but strong contrasts require effective preconditioners to obtain a convergent iteration [12, 25]. The FFT-accelerated evaluation of the Green operator that we adopt in this study relies on the same classical construction where we use it to assemble the action of the operator on the trial space spanned by the plane-wave neural basis functions. This yields a differentiable solver for the scattered field that can be evaluated at arbitrary spatial locations.

Trefftz methods, plane-wave basis, and Herglotz wave functions. Trefftz and quasi-Trefftz methods use basis functions that satisfy the governing equation exactly, or up to a high-order local error, rather than generic polynomials. Generalized plane-wave methods extend this idea to smoothly varying media [17]. Our plane-wave neural basis is related in spirit: the plane-wave carriers solve the homogeneous Helmholtz equation exactly, while the neural amplitude captures the spatially varying correction. Our approach differs from plane-wave activation networks [10], in which plane-waves are used as the activation functions of a single-hidden-layer network, whereas we combine plane-wave carriers with a deep neural amplitude and target general heterogeneous, spatially varying wavenumber media rather than a homogeneous background. It also differs from the plane-wave discontinuous Galerkin domain-decomposition method of [31], which is formulated for piecewise-constant wavenumbers, whereas our method addresses smoothly and strongly varying velocity models within a single global trial space.

Our approach is based on Herglotz wave functions, which superpose plane-waves at a fixed wavenumber and integrated over all directions on the unit sphere. Classical results show such superpositions are dense in the solution space of the homogeneous Helmholtz equation, with approximation properties in Sobolev norms studied rigorously [8]. Each basis function carries a plane wave at a fixed direction, sampled from a finite set of directions on the unit sphere, together with a trainable wavenumber and a trained amplitude that modulates it. The homogeneous solution space is thus approximated over a finite, discrete set of directions by varying wavenumber and amplitude, rather than over a continuum of directions at a single fixed wavenumber. Classical density results for plane-wave bases, such as [8], are stated for a continuum of directions at fixed wavenumber, so they do not apply to our construction. We prove a separate density result for our finite-direction, trainable-wavenumber basis in Theorem 3.3.

Neural solvers for Helmholtz and seismic modeling. PINNs and related neural approaches have been applied to frequency-domain wave modeling in heterogeneous media, including Gabor, Fourier-feature, and Green-integral-constrained formulations [1, 2, 3, 11, 29]. These methods train the network end-to-end on the Helmholtz equation or its integral formulation, relying solely on the network’s own representational capacity, without introducing an inductive bias such as the plane-wave basis proposed here. Oscillatory structures are generally incorporated into the network architecture through input encoding techniques, such as positional encoding [1], or by warm-starting the training from models previously trained at lower frequencies [16]. In contrast, our approach embeds this structure directly in the trial space through the plane-wave carriers, which operate effectively on the physical coordinates without any feature embedding.

Least-squares for the linear coefficients. The idea of adaptive basis functions with a linear output layer solved by least squares appears independently in several lines of work. Extreme learning machines fix a single hidden layer at random and solve only for the output weights by least squares [15]. The variable projection (VarPro) principle for nonlinear least-squares problems formalizes this split more generally: the linear parameters are solved using least squares, and the first-order optimization is performed over the nonlinear parameters alone [13]. This principle has been adopted in PINNs (see e.g., [3, 5]), where the network parameters are optimized by Adam while the weights of the last layer are solved exactly by least squares. Our approach differs in that all neural basis functions, taken from every hidden layer rather than only the last, are connected to the output, treating each neuron as an individual basis function. We then use the exact least-squares solve for the coefficients coupled with gradient-based training of the nonlinear network parameters. Unlike a single hidden layer, a deep network builds hierarchical features across layers, so treating every hidden neuron as a basis function lets us capture local, fine-scale structure in the field that a shallow basis cannot represent.

Contributions and paper organization

We summarize our contributions as follows.

  • A plane-wave neural basis for time-harmonic wave propagation, in which every neuron of every hidden layer contributes a basis function modulated by an oscillatory plane-wave carrier with trainable wavenumber and amplitude.

  • An extension of the variable projection scheme to all layers, in which the linear coefficients are obtained at each step from a regularized least-squares solve, while gradient-based training updates only the nonlinear carrier and amplitude parameters.

  • An adaptive-depth training strategy that grows the architecture layer-by-layer instead of fixing it at full depth from the start. Combined with the exact coefficient solve, it removes the loss spikes that occur when a new layer is added and keeps the error decreasing smoothly, an effect that neither ingredient produces alone.

  • A GPU implementation in JAX combining an FFT evaluation of the Green operator with a complex-valued direct solve in double precision. It reaches relative L2L^2 errors of order 1% against exact 2D and 3D solutions, and 5%, 9%, and 10% against the reference solution on Marmousi (10 Hz), Overthrust (10 Hz), and Otway (20 Hz), with the 10-Hz Marmousi case solved in roughly three minutes on a single GPU.

The remainder of the paper is organized as follows. In Section 2, we recall the Lippmann-Schwinger equation and its abstract properties. In Section 3, we present the proposed methodology. In Section 4, we describe the algorithmic and computational aspects of the proposed method. In Section 5, we describe the benchmark problems and the reference solutions. In Section 6, we report the numerical experiments. In Section 7, we summarize the main conclusions of this study. Further details are given in Appendices B-F.

Mathematical model: The Lippmann-Schwinger equation

In this section, we recall the Lippmann-Schwinger equation, the mathematical model adopted in this study to describe time-harmonic acoustic wave propagation in heterogeneous media.

Setting and medium contrast

Let Ω⊂Rd\Omega\subset\mathbb{R}^{d} be a bounded connected open set with d=2,3d=2,3. We consider time-harmonic wave propagation at angular frequency ω=2πf0>0\omega=2\pi f_{0}>0, where f0>0f_{0}>0 denotes the temporal frequency. The wave speed is denoted by v:Ω→Rv:\Omega\to\mathbb{R} and we assume that there exist constants 0<vmin⁡≤vmax⁡<∞0<v_{\min}\le v_{\max}<\infty such that vmin⁡≤v(x)≤vmax⁡v_{\min}\le v(x)\le v_{\max} for all x∈Ωx\in\Omega. The local wavenumber is given by k(x)=ω/v(x)k(x)=\omega/v(x), and we denote the squared slowness by m(x)=v(x)−2m(x)=v(x)^{-2}. The corresponding background squared slowness and wavenumber are m0=v0−2m_{0}=v_{0}^{-2} and k0=ω/v0k_{0}=\omega/v_{0}, respectively, where v0>0v_{0}>0 is the constant background wave speed. The medium contrast, which quantifies the deviation of the true medium from the background, is defined by

δm(x)=m(x)−m0,(1)\delta m(x)=m(x)-m_{0}, \tag*{(1)}

and satisfies δm∈L∞(Ω)\delta m\in L^{\infty}(\Omega) with supp⁡(δm)⊂Ω\operatorname{supp}(\delta m)\subset\Omega, which ensures that the scatterer is compactly supported within Ω\Omega.

Incident and scattered fields

The total acoustic field uu is decomposed as

u=u0+us,(2)u=u_{0}+u_{s}, \tag*{(2)}

where u0u_{0} denotes the incident field, propagating in a prescribed homogeneous background medium with constant wave speed v0>0v_{0}>0, and usu_{s} is the scattered field arising from the medium heterogeneities.

Let G0G_{0} denote the outgoing free-space Green’s function of the background Helmholtz operator L0:=−Δ−ω2m0L_{0}:=-\Delta-\omega^{2}m_{0}, that is, the unique solution of L0G0(⋅,y)=δyL_{0}G_{0}(\cdot,y)=\delta_{y} satisfying the Sommerfeld radiation condition at infinity. Since v0v_{0} is constant, G0G_{0} is translation invariant and admits the explicit kernel representation

G0(x,y)=g0(x−y),x≠y.(3)G_{0}(x,y)=g_{0}(x-y),\qquad x\ne y. \tag*{(3)}

In this study, we consider the free-space Green’s function

g0(x)={i4H0(2)(k0∥x∥),d=2,exp⁡(−ik0∥x∥)4π∥x∥,d=3,(4)g_{0}(x)= \begin{cases} \frac{\mathrm{i}}{4}H_{0}^{(2)}(k_{0}\lVert x\rVert), & d=2,\\ \frac{\exp(-\mathrm{i}k_{0}\lVert x\rVert)}{4\pi\lVert x\rVert}, & d=3, \end{cases} \tag*{(4)}

where H0(2)H_{0}^{(2)} is the Hankel function of the second kind and order zero. For a point source located at xs∈Ωx_{s}\in\Omega, the incident field is accordingly the outgoing solution of the homogeneous background problem,

u0(x)=g0(x−xs).(5)u_{0}(x)=g_{0}(x-x_{s}). \tag*{(5)}

The Green’s integral operator and the governing equation

The Green’s integral operator G:L2(Ω)→L2(Ω)\mathcal{G}:L^{2}(\Omega)\to L^{2}(\Omega) is defined, for φ∈L2(Ω)\varphi\in L^{2}(\Omega) and x∈Ωx\in\Omega, by

(Gφ)(x):=ω2∫Ωg0(x−y) δm(y) φ(y) dy.(6)(\mathcal{G}\varphi)(x):=\omega^{2}\int_{\Omega}g_{0}(x-y)\,\delta m(y)\,\varphi(y)\,\mathrm{d}y. \tag*{(6)}

The scattered field usu_{s} we seek must satisfy the Lippmann-Schwinger equation: given an incident field u0∈L2(Ω)u_{0}\in L^{2}(\Omega), find us∈L2(Ω)u_{s}\in L^{2}(\Omega) such that

(I−G)us=Gu0,(7)(I-\mathcal{G})u_{s}=\mathcal{G}u_{0}, \tag*{(7)}

where II denotes the identity operator on L2(Ω)L^{2}(\Omega).

Abstract formulation of the Lippmann-Schwinger equation

This section outlines the abstract formulation of the Lippmann-Schwinger equation. These theoretical foundations are well-established within scattering theory, and a comprehensive mathematical treatment is available in [20].

The Green’s operator defined in (6) is a linear map G:L2(Ω)→L2(Ω)\mathcal{G}:L^2(\Omega)\to L^2(\Omega) that possesses several mathematical properties under suitable, physically realistic assumptions. First, Lemma C.1 (Appendix C) establishes that G\mathcal{G} is a Hilbert-Schmidt operator, which in particular implies that it is compact and bounded. Moreover, G\mathcal{G} maps L2(Ω)L^2(\Omega) into Hloc2(Rd)H^2_{\mathrm{loc}}(\mathbb{R}^d), where the local Sobolev space is defined by

Hloc2(Rd)={u:Rd→C∣u∣K∈H2(K) for every bounded open set K⊂Rd}.(8)H^2_{\mathrm{loc}}(\mathbb{R}^d)=\{u:\mathbb{R}^d\to\mathbb{C}\mid u|_K\in H^2(K)\text{ for every bounded open set }K\subset\mathbb{R}^d\}. \tag*{(8)}

Consequently, G\mathcal{G} acts as a smoothing operator that elevates the spatial regularity by two orders of differentiation.

The Lippmann-Schwinger equation (7) is formulated as follows. Given a source term f=Gu0∈L2(Ω)f=\mathcal{G}u_0\in L^2(\Omega), we seek the unique scattered field us∈L2(Ω)u_s\in L^2(\Omega) satisfying

Bus=f,(9)\mathcal{B}u_s=f, \tag*{(9)}

where B=I−G\mathcal{B}=\mathcal{I}-\mathcal{G}. The following proposition shows that B\mathcal{B} is boundedly invertible on L2(Ω)L^2(\Omega), its proof, which follows the classical Fredholm-alternative and unique-continuation argument of [20], is given in Appendix C.§ 1.

Proposition 2.1 (Well-posedness of the scattering operator). Let Ω⊂Rd\Omega\subset\mathbb{R}^d be a bounded connected open set with d=2,3d=2,3, and let δm∈L∞(Ω)\delta_m\in L^\infty(\Omega). Let G0(x,y)G_0(x,y) be the outgoing free-space Green’s function of the background operator L0=−Δ−ω2m0L_0=-\Delta-\omega^2m_0, where k0=ω/v0k_0=\omega/v_0. Then the operator B=I−G:L2(Ω)→L2(Ω)\mathcal{B}=\mathcal{I}-\mathcal{G}:L^2(\Omega)\to L^2(\Omega) is boundedly invertible.

Approximation space and training methodology

In this study, our main objective is the accurate approximation of the scattered wavefield usu_s in heterogeneous media for realistic benchmark scenarios. We develop a data-free numerical approach that employs a NN ansatz to approximate the unknown field usu_s, trained in an unsupervised manner directly by enforcing the Lippmann-Schwinger equation (7).

Residual minimization

Since closed-form exact solutions of heterogeneous media scattering problems are generally unavailable, we seek an approximate solution via a residual minimization method. Recall that the scattered field us∈L2(Ω)u_s\in L^2(\Omega) satisfies the Lippmann-Schwinger equation Bus=f\mathcal{B}u_s=f, where B=I−G\mathcal{B}=\mathcal{I}-\mathcal{G} and G\mathcal{G} is the integral operator defined in (6). Given an incident source f∈L2(Ω)f\in L^2(\Omega), we introduce the least-squares functional J:L2(Ω)→R≥0J:L^2(\Omega)\to\mathbb{R}_{\geq0} defined by

J(v)=12∥Bv−f∥L2(Ω)2.(10)J(v)=\frac{1}{2}\lVert\mathcal{B}v-f\rVert_{L^2(\Omega)}^2. \tag*{(10)}

As shown in Theorem 3.1, JJ is a strictly convex quadratic functional admitting a unique global minimizer, which coincides with the unique solution of the governing integral equation; the proof is given in Appendix C.§ 2.

Theorem 3.1 (Existence and uniqueness of the global minimizer). The least-squares functional J:L2(Ω)→R≥0J:L^2(\Omega)\to\mathbb{R}_{\geq0} given by (10) is strictly convex and quadratic, and admits a unique global minimizer us∈L2(Ω)u_s\in L^2(\Omega), which is the unique solution of Bus=f\mathcal{B}u_s=f.

A central question is whether minimizing this residual-based loss also controls the true approximation error, e.g., in the L2L^2 sense. We show that this is indeed the case and therefore, minimizing JJ is equivalent to minimizing the true error, as stated precisely in the following theorem; see Appendix C.§ 3 for the proof.

References

  1. [1]M. M. Abedi, D. Pardo, and T. Alkhalifah. Gabor-enhanced physics-informed neural networks for fast simulations of acoustic wavefields. Neural Networks, 193:107978, 2026. ISSN 0893-6080. doi: https://doi.org/10.1016/j.neunet.2025.107978.
  2. [2]M. M. Abedi, D. Pardo, and T. Alkhalifah. A Green-Integral–Constrained Neural Solver With Stochastic Physics-Informed Regularization. IEEE Transactions on Geoscience and Remote Sensing, 64:5917014–5917014, 2026. doi: 10.1109/TGRS.2026.3720567.DOI
  3. [3]M. M. Abedi, D. Pardo, and T. Alkhalifah. Least-squares-embedded optimization for accelerated convergence of PINNs in high-frequency acoustic wavefield simulations. Computers & Geosciences, page 106162, 2026.DOI
  4. [4]F. Aminzadeh, N. Burkhard, L. Nicoletis, F. Rocca, and K. Wyatt. SEG/EAEG 3-D modeling project: 2nd update. The Leading Edge, 13(9):949–952, 1994.DOI
  5. [5]S. Baharlouei, J. M. Taylor, and D. Pardo. A Least-Squares-Based Regularity-Conforming Neural Network (LS-ReCoNN) for Solving Parametric Transmission Problems. Computer Methods in Applied Mechanics and Engineering, 461:119206, 2026.DOI
  6. [6]A. Berkhout. Combining full wavefield migration and full waveform inversion, a glance into the future of seismic imaging. Geophysics, 77(2):S43–S50, 2012.DOI
  7. [7]A. Brougois, M. Bourget, P. Lailly, M. Poulet, P. Ricarte, and R. Versteeg. Marmousi, model and data. In EAEG workshop-practical aspects of seismic data inversion, pages cp–108. European Association of Geoscientists & Engineers, 1990.DOI
  8. [8]D. Colton and R. Kress. On the denseness of Herglotz wave functions and electromagnetic Herglotz pairs in Sobolev spaces. Mathematical methods in the applied sciences, 24(16):1289–1303, 2001.DOI
  9. [9]M. Costabel, E. Darrigrand, and H. Sakly. The essential spectrum of the volume integral operator in electromagnetic scattering by a homogeneous body. Comptes Rendus Mathématique, 350:193–197, 2012.DOI
  10. [10]T. Cui, Z. Wang, and X. Xiang. An efficient neural network method with plane wave activation functions for solving Helmholtz equation. Computers & Mathematics with Applications, 111:34–49, 2022.DOI
  11. [11]Y. Ding, S. Chen, H. Miyake, and X. Li. Physics-informed neural networks with Fourier features for seismic wavefield simulation in time-domain nonsmooth complex media. IEEE Transactions on Geoscience and Remote Sensing, 63:1–13, 2025.
  12. [12]K. S. Eikrem, G. Næv̊dal, and M. Jakobsen. Iterative solution of the Lippmann–Schwinger equation in strongly scattering acoustic media by randomized construction of preconditioners. Geophysical Journal International, 224 (3):2121–2130, 03 2021. doi: 10.1093/gji/ggaa503. URL https://doi.org/10.1093/gji/ggaa503.
  13. [13]G. H. Golub and V. Pereyra. The differentiation of pseudo-inverses and nonlinear least squares problems whose variables separate. SIAM Journal on numerical analysis, 10(2):413–432, 1973.DOI
  14. [14]L. Guasch, O. Calderón Agudo, M.-X. Tang, P. Nachev, and M. Warner. Full-waveform inversion imaging of the human brain. NPJ digital medicine, 3(1):28, 2020.DOI
  15. [15]G.-B. Huang, Q.-Y. Zhu, and C.-K. Siew. Extreme learning machine: theory and applications. Neurocomputing, 70(1-3):489–501, 2006.DOI
  16. [16]X. Huang and T. Alkhalifah. PINNup: Robust neural network wavefield solutions using frequency upscaling and neuron splitting. Journal of Geophysical Research: Solid Earth, 127(6):e2021JB023703, 2022.DOI
  17. [17]L.-M. Imbert-Gérard and B. Després. A generalized plane-wave numerical method for smooth nonconstant coefficients. IMA Journal of Numerical Analysis, 34(3):1072–1103, 2014.
  18. [18]M. Jakobsen, X. Huang, and R.-S. Wu. Homotopy analysis of the Lippmann–Schwinger equation for seismic wavefield modelling in strongly scattering media. Geophysical Journal International, 222(2):743–753, 2020.DOI
  19. [19]D. P. Kingma and J. Ba. Adam: A Method for Stochastic Optimization. In 3rd International Conference on Learning Representations (ICLR), 2015.arxiv.org/abs/1412.6980
  20. [20]A. Kirsch. An introduction to the mathematical theory of inverse problems, volume 120. Springer, 2011.
  21. [21]C. Krishnanunni, T. Bui-Thanh, and C. Dawson. Topological derivative approach for deep neural network architecture adaptation. arXiv preprint arXiv:2502.06885, 2025.arxiv.org/abs/2502.06885
  22. [22]A. Lechleiter and P. Monk. The time-domain Lippmann–Schwinger equation and convolution quadrature. Numerical Methods for Partial Differential Equations, 31(2):517–540, 2015. doi: https://doi.org/10.1002/num. 21921.
  23. [23]P. M. Morse and H. Feshbach. Methods of theoretical physics. Technology Press, 1946.
  24. [24]S. Nair, T. F. Walsh, G. Pickrell, and F. Semperlotti. Multiple scattering simulation via physics-informed neural networks. Engineering with Computers, 41(1):31–50, 2025.DOI
  25. [25]G. Osnabrugge, S. Leedumrongwatthanakun, and I. M. Vellekoop. A convergent Born series for solving the inhomogeneous Helmholtz equation in arbitrarily large media. Journal of computational physics, 322:113–124, 2016.DOI
  26. [26]P. Petersen and J. Zech. Mathematical theory of deep learning. arXiv preprint arXiv:2407.18384, 2024.arxiv.org/abs/2407.18384
  27. [27]B. Ronen, D. Jacobs, Y. Kasten, and S. Krtichman. The convergence rate of neural networks for learned functions of different frequencies. Advances in Neural Information Processing Systems, 32, 2019.arxiv.org/abs/1906.00425
  28. [28]J. H. Ryoo and C.-H. Park. Lippmann-Schwinger approach for accurate photoelectron wave functions and angle-resolved photoemission spectra from first principles. Physical Review Letters, 135(5):056403, 2025.DOI
  29. [29]C. Song, T. Alkhalifah, and U. B. Waheed. Solving the frequency-domain acoustic VTI wave equation using physics-informed neural networks. Geophysical Journal International, 225(2):846–859, 2021.DOI
  30. [30]W. Wen, F. Yan, Y. Chen, and H. Li. Autogrow: Automatic layer growing in deep convolutional networks. In Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pages 833–841, 2020.arxiv.org/abs/1906.02909
  31. [31]L. Yuan and Q. Hu. A discontinuous plane wave neural network method for Helmholtz equation and time-harmonic Maxwell’s equations. Advances in Computational Mathematics, 51(2):18, 2025.arxiv.org/abs/2310.09527

Paper details

vlnc:W_s4nd3e99
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 2

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents