The reviewed record of science sign in
Pith

arxiv: 2607.06274 · v1 · pith:EHVLRABX · submitted 2026-07-07 · cs.SD · eess.AS

Learning-based Physics-Constrained Neural Kernel for Sound Field Estimation With Source-Position-Dependent Directional Weighting

Mattia Marella , Shoichi Koyama This is my paper

Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel 2026-07-08 11:09 UTCglm-5.2pith:EHVLRABXrecord.jsonopen to challenge →

classification cs.SD eess.AS
keywords sound field estimationkernel regressionimplicit neural representationphysics-constrained neural kerneldirectional weightingHerglotz wave functionacoustic transfer functionroom acoustics
0
0 comments X

The pith

Neural kernel learns room acoustics from many sources, not one

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

The paper addresses the problem of estimating the spatial distribution of a sound field from a small set of microphone measurements. Prior physics-constrained neural kernel methods adapt a directional weighting function to a single snapshot of microphone data, which overfits and fails to generalize. The authors propose making the directional weighting function an implicit neural representation that takes both the propagation direction and the source position as inputs. By training this network on acoustic transfer functions from many source positions in a room, the kernel learns common directional patterns—such as direct sound and early reflection directions—and produces an appropriate weighting function for unseen source positions without requiring per-measurement optimization. Experiments on simulated room acoustics show lower reconstruction error and tighter error distributions compared to the snapshot-based approach, with the learned weighting function aligning to true source and image-source directions.

Core claim

The central mechanism is a source-position-dependent implicit neural representation (INR) for the directional weighting function inside a Herglotz-wave kernel. The Herglotz wave function expresses the acoustic field as a weighted integral of plane waves over the unit sphere; the weighting function determines which propagation directions the kernel emphasizes. By parameterizing this weighting as a neural network that conditions on both the propagation direction and the source position, and by training the network on transfer functions from multiple sources in a fixed room, the model learns a mapping from source position to directional pattern. At inference time, given a new source position, a

What carries the argument

Herglotz wave function kernel, implicit neural representation (INR), Random Fourier Features, Lebedev quadrature, image source method

If this is right

  • If the source-position-dependent INR generalizes beyond the tested shoebox room, it could enable pre-computed acoustic models for specific venues that produce accurate sound field estimates from sparse microphones without per-event calibration.
  • The multi-source training paradigm could be extended to multiple rooms, potentially learning a universal mapping from room geometry and source position to directional weighting, as the authors note in their conclusion.
  • The learned weighting functions visually align with image-source directions, suggesting the INR implicitly discovers the room's reflection structure, which could be extracted for room geometry inference or acoustic analysis.
  • Since inference reduces to a forward pass through the INR plus a linear kernel operation, the approach is compatible with real-time spatial audio applications.

Where Pith is reading between the lines

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

  • The alignment of learned weighting functions with image-source directions suggests the INR may be implicitly encoding a compressed representation of the room's image-source geometry, raising the question of whether explicit geometric priors could improve sample efficiency.
  • Training on 80 source positions in a single simulated room with moderate reverberation (T60=200ms) may not capture the diversity of real acoustic environments; generalization to measured data with diffuse late reverberation, non-convex geometries, or higher reverberation times remains the critical untested boundary.
  • The bandwidth scaling of the Random Fourier Features proportional to the wavenumber implies that higher frequencies require higher-capacity embeddings, which could create a computational bottleneck for broadband or high-frequency applications.

Load-bearing premise

The claim of generalization rests on tests within a single simulated shoebox room with moderate reverberation; whether the learned directional weighting function transfers to real measured data, different room shapes, or more reverberant environments is not established.

What would settle it

If the source-position-dependent INR, trained on synthetic transfer functions from one room, produces reconstruction errors no better than the snapshot-based method when evaluated on measured acoustic transfer functions or on rooms with different geometries and reverberation times, the central claim of generalization would not hold.

read the original abstract

A learning-based physics-constrained neural kernel for sound field estimation is proposed. Sound field estimation aims to estimate the spatial distribution of an acoustic field from a discrete set of microphone measurements, which have a wide range of applications. Among existing sound field estimation methods, kernel-regression-based methods offer a flexible and principled framework for incorporating physical constraints and allow inference through linear operation. It is also possible to adapt the kernel function to the target acoustic environment by representing the directional weighting function as an implicit neural representation (INR) and optimizing hyperparameters using measurements. However, the kernel function is generally optimized for single snapshot measurements of the microphones, which can lead to strong overfitting and poor generalization. We propose a source-position-dependent INR for the directional weighting function, enabling the kernel function to capture common directional patterns and to generalize to unseen source positions in the target acoustic environment. Experimental results indicate that our proposed method outperforms the snapshot-based method by estimating a directional weighting function that matches the directivity of the target sound field.

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 / 7 minor

Summary. This paper proposes a learning-based physics-constrained neural kernel (LB-NK) for sound field estimation. The key idea is to represent the directional weighting function in the Herglotz-wave-function-based kernel as an implicit neural representation (INR) that depends on the source position, trained on a set of acoustic transfer functions (ATFs) from multiple sources in a fixed room. This allows the kernel to generalize to unseen source positions without per-snapshot fine-tuning, addressing the overfitting problem of the snapshot-based neural kernel (SB-NK) of Ribeiro et al. [18]. The Helmholtz constraint is preserved by construction through the kernel formulation. Experiments on synthetic image-source-method data in a single shoebox room show that LB-NK outperforms SB-NK and a uniform-weighting baseline in normalized mean square error (NMSE) across frequencies.

Significance. The paper makes a reasonable contribution to physics-informed sound field estimation by extending the neural kernel framework from single-snapshot optimization to multi-source training. The mathematical framework is clean: the Herglotz wave function kernel (Eq. 4) preserves the Helmholtz constraint by construction, and the RKHS estimation (Eq. 3) is standard. The source-position-dependent INR is a natural and well-motivated extension. The Softplus output guarantees positive semi-definiteness of the kernel. The directional weighting visualizations in Fig. 6 provide qualitative evidence that the learned weights align with physical source and image-source directions. However, the experimental evaluation is limited to a single simulated room, and the comparison protocol between LB-NK and SB-NK has a methodological asymmetry that needs to be addressed.

major comments (2)
  1. §5.1, experimental design: LB-NK benefits from validation-based early stopping, while SB-NK is denied any validation mechanism and uses a fixed 300-epoch budget. This asymmetry confounds the central comparison. The paper's own motivation states that single-snapshot optimization 'can lead to strong overfitting'; if 300 epochs causes SB-NK to overfit, the reported performance gap may partly reflect the absence of early stopping rather than the benefit of multi-source training alone. A fairer comparison would provide SB-NK with a validation mechanism, e.g., by holding out a subset of the M microphones as a validation set for model selection. Without this control, the claim that LB-NK outperforms SB-NK due to source-position-dependent training is not fully isolated from the benefit of having a validation set.
  2. §5.1–5.2, scope of evaluation: The entire experimental evaluation uses a single simulated shoebox room (T60=200 ms) with image-source-method data. The claim that the learned INR 'generalizes to unseen source positions' (Abstract, §4.1) is tested only within this narrow setting. Whether the learned directional weighting generalizes to real measured ATFs, non-shoebox geometries, or different reverberation times is unaddressed. The conclusion mentions 'future work to generalize to multiple rooms,' but the current scope is quite limited for a method whose central selling point is generalization. At minimum, the paper should discuss this limitation explicitly and clarify that the generalization claim is currently restricted to unseen sources within the same simulated room.
minor comments (7)
  1. §4.2, Eq. (6): The RFF bandwidth B is stated to be 'scaled proportionally to k,' but the proportionality constant is not given. Please specify.
  2. §5.1: The regularization parameter λ in Eq. (3) is not stated. Was it the same for LB-NK, SB-NK, and Uniform? Please specify the value and whether it was tuned.
  3. §5.1: The number of Lebedev quadrature points D is not stated. Please state the value used.
  4. Fig. 4: The frequency axis labels are difficult to read. Consider enlarging or providing a table of key values.
  5. Fig. 5: The boxplot frequency labels are small. Consider enlarging or annotating directly on the figure.
  6. §5.2: The claim that LB-NK's tighter inter-quartile ranges indicate a 'structural regularizer' effect would be strengthened by reporting the number of test sources and whether the differences are statistically significant.
  7. §3.2: The directed component (von Mises–Fisher superposition) from [18] is described, but it is unclear whether LB-NK also uses this directed component or only the INR. Please clarify the full LB-NK architecture used.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and constructive feedback. The referee raises two major points: (1) a methodological asymmetry in the comparison between LB-NK and SB-NK regarding validation-based early stopping, and (2) the limited experimental scope (single simulated shoebox room). We agree that both points warrant attention and will revise the manuscript accordingly. For the first point, we will add an additional SB-NK baseline that uses a held-out microphone subset for validation-based early stopping, to isolate the benefit of multi-source training from the benefit of having a validation set. For the second point, we will explicitly acknowledge the scope limitation of the generalization claim and clarify that it is currently restricted to unseen sources within the same simulated room.

read point-by-point responses

  1. Referee: LB-NK benefits from validation-based early stopping, while SB-NK is denied any validation mechanism and uses a fixed 300-epoch budget. This asymmetry confounds the central comparison. A fairer comparison would provide SB-NK with a validation mechanism, e.g., by holding out a subset of the M microphones as a validation set for model selection.

    Authors: The referee is correct that the current comparison has a methodological asymmetry: LB-NK uses validation-based early stopping while SB-NK uses a fixed 300-epoch budget without any validation mechanism. We agree that this confounds the comparison and that a fairer protocol is needed to isolate the benefit of multi-source training from the benefit of having a validation set. We will address this in the revision by adding an additional SB-NK baseline that uses a held-out subset of microphones for validation-based early stopping. Specifically, we will hold out a subset of the M microphones as a validation set, optimize SB-NK on the remaining microphones, and use the validation loss for model selection. This will allow us to disentangle the contribution of the source-position-dependent INR from the contribution of early stopping. We note that even with this control, SB-NK still optimizes the directional weighting for a single snapshot without any mechanism to share information across source positions, so the core architectural advantage of LB-NK remains. However, we agree that the current presentation does not fully isolate this advantage, and the revised experiments will make the comparison fairer. revision: yes

  2. Referee: The entire experimental evaluation uses a single simulated shoebox room (T60=200 ms) with image-source-method data. The claim that the learned INR 'generalizes to unseen source positions' is tested only within this narrow setting. Whether the learned directional weighting generalizes to real measured ATFs, non-shoebox geometries, or different reverberation times is unaddressed. At minimum, the paper should discuss this limitation explicitly and clarify that the generalization claim is currently restricted to unseen sources within the same simulated room.

    Authors: The referee is correct that the experimental evaluation is limited to a single simulated shoebox room with T60=200 ms, and that the generalization claim is currently tested only within this narrow setting. We agree that the abstract and the body of the paper should be more precise about the scope of the generalization claim. In the revision, we will explicitly state that the current evaluation demonstrates generalization to unseen source positions within the same simulated acoustic environment, and that generalization across different rooms, geometries, and reverberation times is left as future work. We will add this clarification to the abstract, Section 5.1, and the conclusion. Regarding additional experiments with real measured data or different room configurations: while we agree these would strengthen the paper, we are constrained by the availability of suitable multi-source ATF datasets with dense ground-truth evaluation points. We will, however, add a discussion of this limitation and outline the specific challenges and planned directions for extending the evaluation to real and diverse environments. revision: partial

Circularity Check

0 steps flagged

No significant circularity: the kernel construction is structural (Helmholtz-constrained by definition), the INR is trained on training sources and evaluated on held-out test sources, and the central comparison is not circular by construction.

full rationale

The paper's core derivation is not circular. The kernel function (Eq. 4) is defined via the Herglotz wave function with a directional weighting w(η), and the Helmholtz constraint is structural—any field estimated via Eq. (3) with this kernel satisfies the Helmholtz equation by construction of the RKHS, not by fitting. The directional weighting w(η) = INR(η, y; θ) (Eq. 5) is trained on ATFs from 80% of sources and evaluated on a held-out 10% test set; the test loss (Eq. 7) is computed on sources not seen during training, so the 'generalization to unseen source positions' claim is not tautological. The comparison with SB-NK involves an experimental asymmetry (LB-NK gets validation-based early stopping, SB-NK gets fixed 300 epochs), but this is a methodological confound affecting the fairness of the comparison, not a circularity in the derivation chain. The only mild self-citation is to [18] (Ribeiro et al., 2024), co-authored by Koyama, which provides the neural kernel framework being extended; this is standard building-on-prior-work, not a load-bearing circular dependency, since the present paper's contribution (adding source-position dependence y to the INR input) is an independent architectural modification evaluated against external benchmarks. No step reduces to its inputs by definition or fit.

Axiom & Free-Parameter Ledger

4 free parameters · 4 axioms · 0 invented entities

The paper introduces no new physical entities, particles, forces, or dimensions. The INR and neural kernel are mathematical constructs within an established framework. The free parameters are standard machine learning hyperparameters and the RKHS regularization parameter. The primary concern is the unstated value of λ and the qualitative description of the RFF bandwidth.

free parameters (4)
  • θ (NN weights) = Learned via Adam optimization
    The INR parameters are optimized to minimize the NMSE loss over training ATFs. These are the primary fitted parameters.
  • λ (regularization parameter in Eq. 3) = Not stated in the paper
    The Tikhonov regularization parameter in the kernel ridge regression solution is used but its value is never specified, making it an unstated free parameter.
  • B (RFF random matrix bandwidth) = Scaled proportionally to k
    The Gaussian random matrix for Random Fourier Features has a bandwidth described as proportional to the wavenumber k, but the exact proportionality constant is not given.
  • D (number of Lebedev quadrature points) = Not explicitly stated
    The number of discrete directions for numerical integration of the kernel is not specified, though the Lebedev quadrature method is cited.
axioms (4)
  • domain assumption The acoustic pressure field satisfies the homogeneous Helmholtz equation in the source-free region of interest (Eq. 1).
    Standard physics assumption for single-frequency sound field estimation in a source-free region.
  • domain assumption Acoustic transfer functions from multiple source positions in a fixed room share common directional patterns that can be learned and generalized to unseen source positions.
    This is the core premise of the learning-based approach (§4.1). It assumes that the directional weighting function's dependence on source position is smooth enough to be learned from 80 training sources.
  • standard math The representer theorem applies, yielding the closed-form kernel solution in Eq. (3).
    Standard RKHS result cited from Schölkopf et al. [24].
  • domain assumption Lebedev quadrature provides sufficient accuracy for the numerical integration of the Herglotz wave function kernel.
    The kernel (Eq. 4) is computed by discretizing S² using Lebedev quadrature (§4.1), but no error analysis of this approximation is provided.

pith-pipeline@v1.1.0-glm · 11511 in / 2724 out tokens · 444198 ms · 2026-07-08T11:09:49.337042+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

39 extracted references · 39 canonical work pages · 4 internal anchors

  1. [1]

    It can be applied to a wide variety of downstream tasks, for example, acoustic imaging [1], room acous- tic analysis [2], and spatial audio reproduction and control [3, 4]

    INTRODUCTION Sound field estimation/reconstruction/interpolation is a fundamental task in audio signal processing and machine learning, aiming to esti- mate the spatial distribution of an acoustic field from a discrete set of sensor (microphone) observations. It can be applied to a wide variety of downstream tasks, for example, acoustic imaging [1], room ...

    work page

  2. [2]

    PROBLEM FORMULA TION LetΩ⊂R 3 be a simply-connected, source-free region of interest. The acoustic pressure field of angular frequencyωat positionx∈Ω is denoted byu(x, ω), which satisfies the homogeneous Helmholtz equation: ∇2u(x, ω) +k 2u(x, ω) = 0,∀x∈Ω,(1) arXiv:2607.06274v1 [cs.SD] 7 Jul 2026 wherek=ω/cis the wavenumber andcis the speed of sound. Sound ...

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  3. [3]

    They allow for inter- ference through linear operation and require few parameters that have to be set empirically, such as the number of basis functions

    RELA TED WORK We here briefly revisit the kernel-regression-based sound field esti- mation methods, which can be regarded as a generalization of finite- dimensional basis-expansion-based methods. They allow for inter- ference through linear operation and require few parameters that have to be set empirically, such as the number of basis functions. 3.1. Ke...

    work page

  4. [4]

    PROPOSED METHOD We propose a learning-based neural kernel for sound field estima- tion, which extends the physics-constrained neural kernel framework of [18] to a multi-source training paradigm. Whereas the current method fits independent kernel instances for each source position using snapshot measurements, the proposed approach learns a direc- tional we...

    work page

  5. [5]

    Experimental Setup To construct a dataset, synthetic room impulse responses (RIRs) are generated by using the image source method [31]

    EXPERIMENTS 5.1. Experimental Setup To construct a dataset, synthetic room impulse responses (RIRs) are generated by using the image source method [31]. The simulated room is4.0 m×6.0 m×3.0 mshoebox geometry, and the target reverberation time is set toT 60 = 200 ms. The target regionΩis a spherical region of radius0.5 mcentered at(2.0 m,3.0 m,1.5 m) with ...

    work page

  6. [6]

    CONCLUSION We proposed a learning-based physics-constrained neural kernel for sound field estimation. The current method suffers from strong over- fitting and poor generalization as the directional weighting function for the kernel function is optimized based solely on the single snap- shot measurements. By constructing the directional weighting func- tio...

    work page

  7. [7]

    Nearfield acoustic holography: I. theory of generalized holography and the de- velopment of nah,

    J. D. Maynard, E. G. Williams, and Y . Lee, “Nearfield acoustic holography: I. theory of generalized holography and the de- velopment of nah,”J. Acoust. Soc. Amer., vol. 78, no. 4, pp. 1395–1413, 10 1985

    work page 1985

  8. [8]

    Sound-field analysis by plane-wave decomposition using spherical microphone array,

    M. Park and B. Rafaely, “Sound-field analysis by plane-wave decomposition using spherical microphone array,”J. Acoust. Soc. Amer., vol. 118, no. 5, pp. 3094–3103, 2005

    work page 2005

  9. [9]

    Three-dimensional surround sound systems based on spherical harmonics,

    M. A. Poletti, “Three-dimensional surround sound systems based on spherical harmonics,”J. Audio Eng. Soc., vol. 53, no. 11, pp. 1004–1025, 2005

    work page 2005

  10. [10]

    Spatial active noise control based on kernel interpolation of sound field,

    S. Koyama, J. Brunnstr ¨om, H. Ito, N. Ueno, and H. Saruwatari, “Spatial active noise control based on kernel interpolation of sound field,”IEEE/ACM Trans. Audio, Speech, Lang. Process., vol. 29, pp. 3052–3063, 2021

    work page 2021

  11. [11]

    Sound field estimation: Theories and applications,

    N. Ueno and S. Koyama, “Sound field estimation: Theories and applications,”Foundations and Trends® in Signal Pro- cessing, vol. 19, no. 1, pp. 1–98, 2025

    work page 2025

  12. [12]

    E. G. Williams,Fourier Acoustics: Sound Radiation and Nearfield Acoustical Holography, Academic Press, 1999

    work page 1999

  13. [13]

    Room impulse response interpolation us- ing a sparse spatio-temporal representation of the sound field,

    N. Antonello, E. De Sena, M. Moonen, P. A. Naylor, and T. van Waterschoot, “Room impulse response interpolation us- ing a sparse spatio-temporal representation of the sound field,” IEEE/ACM Trans. Audio, Speech, Lang. Process., vol. 25, no. 10, pp. 1929–1941, 2017

    work page 1929

  14. [14]

    Source- location-informed sound field recording and reproduction,

    S. Koyama, K. Furuya, Y . Haneda, and H. Saruwatari, “Source- location-informed sound field recording and reproduction,” IEEE J. Sel. Topics Signal Process., vol. 9, no. 5, pp. 881–894, 2015

    work page 2015

  15. [15]

    Compres- sive sensing in acoustic imaging,

    N. Bertin, L. Daudet, V . Emiya, and R. Gribonval, “Compres- sive sensing in acoustic imaging,” inCompressed Sensing and its Applications, H. Boche, R. Calderbank, G. Kutyniok, and J. Vybiral, Eds. Springer, 2015

    work page 2015

  16. [16]

    Directionally weighted wave field estimation exploiting prior information on source direction,

    N. Ueno, S. Koyama, and H. Saruwatari, “Directionally weighted wave field estimation exploiting prior information on source direction,”IEEE Trans. Signal Process., vol. 69, pp. 2383–2395, 2021

    work page 2021

  17. [17]

    Time-domain sound field es- timation using kernel ridge regression,

    J. Brunnstr ¨om, M. B. Møller, J. Østergaard, S. Koyama, T. van Waterschoot, and M. Moonen, “Time-domain sound field es- timation using kernel ridge regression,”IEEE Trans. Audio, Speech Lang. Process., vol. 34, pp. 1243–1258, 2026

    work page 2026

  18. [18]

    Sound field reconstruction in rooms: Inpainting meets super- resolution,

    F. Llu ´ıs, P. Mart´ınez-Nuevo, M. B. Møller, and S. E. Shepstone, “Sound field reconstruction in rooms: Inpainting meets super- resolution,”J. Acoust. Soc. Amer., vol. 148, no. 2, 2020

    work page 2020

  19. [19]

    Learning neural acoustic fields,

    A. Luo, Y . Du, M. J. Tarr, J. B. Tenenbaum, A. Torralba, and C. Gan, “Learning neural acoustic fields,” inProc. Adv. Neural Inf. Process. Syst. (NeurIPS), 2022

    work page 2022

  20. [20]

    Deep prior approach for room impulse response reconstruction,

    M. Pezzoli, D. Perini, A. Bernardini, F. Borra, F. Antonacci, and A. Sarti, “Deep prior approach for room impulse response reconstruction,”Sensors, vol. 22, no. 7, 2022

    work page 2022

  21. [21]

    Physics-informed convolutional neural network with bicubic spline interpolation for sound field estimation,

    K. Shigemi, S. Koyama, T. Nakamura, and H. Saruwatari, “Physics-informed convolutional neural network with bicubic spline interpolation for sound field estimation,” inProc. Int. Workshop Acoust. Signal Enhancement (IWAENC), Sep. 2022

    work page 2022

  22. [22]

    Reconstruction of sound field through diffusion models,

    F. Miotello, L. Comanducci, M. Pezzoli, A. Bernardini, F. An- tonacci, and A. Sarti, “Reconstruction of sound field through diffusion models,” inProc. IEEE Int. Conf. Acoust., Speech, Signal Process. (ICASSP), Seoul, Republic of Korea, Apr. 2024

    work page 2024

  23. [23]

    Implicit neural rep- resentation with physics-informed neural networks for the re- construction of the early part of room impulse responses,

    M. Pezzoli, F. Antonacci, and A. Sarti, “Implicit neural rep- resentation with physics-informed neural networks for the re- construction of the early part of room impulse responses,” in Proc. Forum Acusticum, Sep. 2023

    work page 2023

  24. [24]

    Sound field estimation based on physics-constrained kernel interpolation adapted to environment,

    J. G. C. Ribeiro, S. Koyama, R. Horiuchi, and H. Saruwatari, “Sound field estimation based on physics-constrained kernel interpolation adapted to environment,”IEEE/ACM Trans. Au- dio, Speech, Lang. Process., vol. 32, pp. 4369–4383, 2024

    work page 2024

  25. [25]

    Room impulse response reconstruction with physics-informed deep learning,

    X. Karakonstantis, D. Caviedes-Nozal, A. Richard, and E. Fernandez-Grande, “Room impulse response reconstruction with physics-informed deep learning,”J. Acoust. Soc. Amer., vol. 155, no. 2, pp. 1048–1059, 2024

    work page 2024

  26. [26]

    Physics-informed neu- ral network for volumetric sound field reconstruction of speech signals,

    M. Olivieri, X. Karakonstantis, M. Pezzoli, F. Antonacci, A. Sarti, and E. Fernandez-Grande, “Physics-informed neu- ral network for volumetric sound field reconstruction of speech signals,”EURASIP J. Audio, Speech, Music Process., , no. 42, 2024

    work page 2024

  27. [27]

    Physics-informed machine learning for sound field estimation: Fundamentals, state of the art, and chal- lenges,

    S. Koyama, J. G. C. Ribeiro, T. Nakamura, N. Ueno, and M. Pezzoli, “Physics-informed machine learning for sound field estimation: Fundamentals, state of the art, and chal- lenges,”IEEE Signal Process. Mag., vol. 41, no. 6, pp. 60–71, 2025

    work page 2025

  28. [28]

    Colton and R

    D. Colton and R. Kress,Inverse Acoustic and Electromagnetic Scattering Theory, Springer, 2013

    work page 2013

  29. [29]

    Implicit neural representations with periodic activation functions,

    V . Sitzmann, J. Martel, A. Bergman, D. Lindell, and G. Wet- zstein, “Implicit neural representations with periodic activation functions,” inProc. Adv. Neural Inf. Process. Syst. (NeurIPS), 2020

    work page 2020

  30. [30]

    A generalized representer theorem,

    B. Sch ¨olkopf, R. Herbrich, and A. J. Smola, “A generalized representer theorem,” inProc. Int. Conf. Comput. Learn. The- ory (COLT), Amsterdam, Netherlands, Jul. 2001, pp. 416–426

    work page 2001

  31. [31]

    K. V . Mardia and P. E. Jupp,Directional Statistics, John Wiley & Sons, 2009

    work page 2009

  32. [32]

    A quadrature formula for the sphere of the 131st algebraic order of accuracy,

    V . I. Lebedev and D. N. Laikov, “A quadrature formula for the sphere of the 131st algebraic order of accuracy,”Doklady Mathematics, vol. 59, pp. 477–481, 1999

    work page 1999

  33. [33]

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

    M. Tancik, P. P. Srinivasan, B. Mildenhall, S. Fridovich-Keil, N. Raghavan, U. Singhal, R. Ramamoorthi, J. T. Barron, and R. Ng, “Fourier features let networks learn high frequency functions in low dimensional domains,” inProc. Adv. Neural Inf. Process. Syst. (NeurIPS), Red Hook, NY , USA, 2020, Cur- ran Associates Inc

    work page 2020

  34. [34]

    Layer Normalization

    J. Ba, J. Kiros, and G. E. Hinton, “Layer normalization,” ArXiv, vol. abs/1607.06450, 2016

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  35. [35]

    Mish: A Self Regularized Non-Monotonic Activation Function

    D. Misra, “Mish: A self regularized non-monotonic neural activation function,”ArXiv, vol. abs/1908.08681, 2019

    work page internal anchor Pith review Pith/arXiv arXiv 1908

  36. [36]

    Incorporating second-order functional knowledge for bet- ter option pricing,

    C. Dugas, Y . Bengio, F. B ´elisle, C. Nadeau, and R. Gar- cia, “Incorporating second-order functional knowledge for bet- ter option pricing,” inProc. Adv. Neural Inf. Process. Syst. (NeurIPS), 2000, pp. 472–478

    work page 2000

  37. [37]

    Image method for efficiently simulating small-room acoustics,

    J. B. Allen and D. A. Berkley, “Image method for efficiently simulating small-room acoustics,”J. Acoust. Soc. Amer., vol. 65, no. 4, pp. 943–950, 1979

    work page 1979

  38. [38]

    Existence of solutions to systems of underdetermined equations and spherical designs,

    X. Chen and R. S. Womersley, “Existence of solutions to systems of underdetermined equations and spherical designs,” SIAM J. Numerical Analysis, vol. 44, no. 6, pp. 2326–2341, 2006

    work page 2006

  39. [39]

    Adam: A Method for Stochastic Optimization

    D. P. Kingma and J. Ba, “Adam: A method for stochastic opti- mization,”CoRR, vol. abs/1412.6980, 2014

    work page internal anchor Pith review Pith/arXiv arXiv 2014