Pith. sign in

REVIEW 3 major objections 6 minor 51 references

PODNO: Proper Orthogonal Decomposition Neural Operators

T0 review · 3 major / 6 minor · reviewed 2026-08-16 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Replacing Fourier modes with data-driven POD modes in a neural operator proves faster and more accurate on high-frequency PDEs.

desk verdict A credible data-adaptive spectral layer for high-frequency neural operators, with a universality theorem that is overclaimed relative to the implemented POD basis. read the letter →

arxiv 2504.18513 v1 pith:273BLQBU submitted 2025-04-25 math.NA cs.LGcs.NAphysics.comp-ph

Zilan Cheng , Zhongjian Wang , Li-Lian Wang , Mejdi Azaiez This is my paper · ORCID
classification math.NAcs.LGcs.NAphysics.comp-ph MSC 68T0765M70
keywords neuraloperatorsproperorthogonaldecompositionFourieroperatordispersivePDEsoscillatorysolutionsuniversalapproximationtheorynonlinearSchrödingerequationKadomtsev-Petviashvili
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper introduces PODNO, a neural operator that builds its spectral kernel on an orthonormal basis computed from the training data by proper orthogonal decomposition (POD) instead of the fixed Fourier basis used by Fourier Neural Operators (FNO). The central claim is that because the POD basis is $L^2$-optimal for compressing the data's energy, the operator learns high-frequency-dominated solutions of dispersive PDEs more accurately than FNO while training and evaluating faster. Experiments on the nonlinear Schrödinger and Kadomtsev–Petviashvili equations show that PODNO achieves lower relative errors than FNO at equal parameter counts, and per-epoch runtimes are shorter. The paper also proves a universality theorem for a generalization called Generalized Spectral Operators (GSO), showing any continuous operator between Sobolev spaces on a compact set can be approximated by such a spectral construction.

What carries the argument

The central object is the POD basis $\Phi_N = \{\phi_k\}_{k=1}^N$ and the induced orthonormal transform $\Pi_N$ that maps a function to its coefficients in this basis. The paper's key move is to use this data-dependent basis inside the spectral convolution layer of FNO, with the learned kernel operator $R_l$ acting on the coefficients and the inverse transform mapping back to physical space. The optimality claim rests on the fact that the POD basis minimizes the $L^2$ projection error for a given number of modes, so the retained modes carry more energy than Fourier modes at the same truncation, which is what lets PODNO keep high-frequency content that FNO discards. The GSO framework isolates the essential property: any orthonormal basis satisfying suitable regularity assumptions yields a universal approximation architecture, with the universality proof adapting the constructive argument used for FNO.

What would settle it

Run PODNO on a high-frequency problem whose snapshots have a slowly decaying POD eigenvalue spectrum (so no small set of modes captures most energy); if its test error then fails to beat FNO's, the energy-optimality mechanism is the bottleneck. Alternatively, check on a concrete dataset whether the computed POD basis includes the constant function and whether the basis functions lie in $H^s$; if either fails, the hypothesis of Theorem 3.1 is violated for that implementation.

Watch

Extended reading notes

Core claim

PODNO takes the FNO architecture and replaces the Fourier transform pair with an orthonormal transform $\Pi_N$ and its inverse built from POD modes of the training snapshots. The kernel integration layer becomes $K_l[v] = \Pi_N^{-1}(R_l \Pi_N v) + W_l[v]$, where $R_l$ is a learned operator on the retained modes. Because the POD basis captures the highest-energy modes of the data under the $L^2$ metric, truncation to $N$ modes preserves more of the solution's energy than truncating the Fourier basis to the same number of modes, which the authors argue is why the operator handles high-frequency oscillatory solutions better. The universality theorem (Theorem 3.1) states that for any continuous operator $G: H^s(\Omega) \to H^{s'}(\Omega)$ with $s' \ge s > d/2$ and any compact subset $K$ of $H^s(\Omega)$, there exists a GSO whose $H^s$ error on $K$ is below any prescribed positive epsilon; PODNO is presented as a special case of GSO. In the numerical comparisons, PODNO achieves lower relative errors and faster runtime than FNO on the NLS and KP equations, while on the Darcy problem FNO retains a small accuracy edge.

Load-bearing premise

The analysis that guarantees universality assumes the chosen basis functions are smooth enough (in the Sobolev space $H^s$) and that the constant function is among them; the POD basis actually used in the experiments is built under the $L^2$ metric and the paper does not show it has these two properties, so the theoretical guarantee may not apply to the implemented algorithm.

Editorial extensions

If this is right

  • On equations with concentrated mode energy, PODNO matches or beats FNO with far fewer modes and parameters: on the KP test, 9 POD modes with 54K parameters reach errors FNO needs 607K parameters to approach.
  • Because the POD transform is a real matrix product rather than a complex FFT, PODNO runs faster per epoch even when parameter counts are equal, and the offline SVD basis cost is small enough to be amortized.
  • The GSO universality theorem implies the FNO-style spectral layer is not tied to the Fourier basis: any sufficiently regular orthonormal basis yields a universal operator architecture.
  • Fourier-domain error analysis shows PODNO's prediction errors stay flat at high modes (L-shaped) while FNO's errors rise (U-shaped), indicating the advantage is specifically in high-frequency content rather than overall smoothness.
  • PODNO's basis construction avoids non-physical padding and the periodic boundary assumption of FFT, so the architecture is not tied to periodic or rectangular domains.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The energy capture ratio $\rho$ computed from snapshots could be used as a practical diagnostic before training: when $\rho$ is small for $N$ modes, PODNO is unlikely to beat FNO, turning a theoretical optimality statement into a testable per-dataset criterion.
  • The GSO viewpoint suggests a broader design space: one could learn the orthonormal basis jointly with the network weights rather than fixing it from snapshots, potentially adapting it to the target operator rather than the marginal data distribution.
  • If a constant mode is deliberately added to the POD basis or the basis functions are regularized into $H^s$, the gap between the implemented algorithm and the universality theorem's assumptions could be closed, making the theoretical guarantee apply to PODNO as run.
  • Because PODNO's transform is a dense matrix multiplication, its runtime advantage over FFT-based FNO may shrink at very large resolutions where FFT's near-linear scaling wins; the current tests at $64 \times 64$ do not probe that regime.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. The paper proposes PODNO, a neural operator architecture that replaces the Fourier transform in FNO's kernel integration with a data-dependent orthonormal transform built from a POD basis. The architecture is embedded in a broader class called Generalized Spectral Operators (GSO), and a universal approximation theorem for GSO is stated and proved constructively. The paper reports numerical experiments on the Darcy problem, the nonlinear Schrödinger equation, and the Kadomtsev–Petviashvili equation, comparing PODNO with FNO under matched parameter counts, and also compares PODNO with a POD-accelerated time-splitting solver. The main claims are that PODNO is more accurate and efficient than FNO on high-frequency-dominated dispersive problems, and that PODNO inherits the universality of GSO.

Significance. If the claims hold, PODNO is a practically interesting alternative to FNO for problems with oscillatory or locally concentrated features, and the GSO universality result extends the existing FNO theory to a wider class of spectral bases. The paper has clear strengths: the numerical comparisons are carefully matched by parameter count, the experiments cover three PDE problems with ablations, and the GSO universality proof is constructive and self-contained. The comparison between PODNO and the POD-accelerated splitting solver is also informative. However, the theoretical result is proved for GSOs built on an H^s-orthonormal basis that must contain the constant function, while the implemented POD basis is only L^2-orthonormal and is explicitly conceded not to satisfy the theorem's assumptions; the conclusion that PODNO itself is universal is therefore not supported as written. The empirical claims would also be stronger with repeated-seed statistics. Overall, the paper contains valuable ingredients, but the central theoretical claim needs either a proof that the POD basis satisfies the required hypotheses or a substantial restriction of the claim.

major comments (3)
  1. [§3.4, Remark 3.1, Appendix A.3] Theorem 3.1 and its proof in Appendix A.3 are formulated for an orthonormal basis in the H^s inner product (Definition A.1) and require the constant function to be one of the basis functions (Assumption 3.3). The implemented PODNO uses an L^2-orthonormal POD basis obtained from data snapshots via SVD (Section 3.2, Algorithm 1). Remark 3.1 concedes that the algorithmic basis is L^2-based, but this is not merely a regularity mismatch: the proof uses P_N as the H^s-orthogonal projection and strong convergence of P_N in H^s to pass to the limit, whereas the implemented POD projection is an L^2 projection onto a data-dependent subspace. No argument is given that this discrete SVD basis provides an H^s-orthonormal basis, contains the constant function, or yields strong H^s convergence of the corresponding projections. Therefore the statement in Section 5 that 'PODNO, as a representative example of GSO, exhibits the same universality' is not supported by Theorem 3.1. The authors should either prove that the L^2-POD basis satisfies the hypotheses of Theorem 3.1 under explicit conditions on the snapshot space, or explicitly restrict the universality claim to GSOs with admissible bases and describe PODNO as an architecture that realizes a GSO only when such a basis is used.
  2. [Assumption 3.1, Proposition A.3, Lemma A.1] The universality proof relies on Proposition A.3, the classical universal approximation theorem for ordinary neural networks, which in the stated form requires a continuous, bounded, non-constant activation function. Assumption 3.1 only assumes a non-polynomial, globally Lipschitz, C^∞ activation and does not require boundedness; the implementation uses GELU, which is unbounded. The proof of Lemma A.1 uses pointwise approximation of Ĝ_N on a compact hypercube, and that step is justified by Proposition A.3 only if the boundedness hypothesis is present. Thus the theorem, as stated, does not cover the activation used in the numerical experiments. The authors should either add boundedness to Assumption 3.1, replace Proposition A.3 with an approximation result valid for unbounded activations, or explicitly state the theorem only for bounded activations.
  3. [§4.3, §4.4, Tables 4.3 and 4.4] The central numerical claim that PODNO outperforms FNO in accuracy and efficiency for the NLS and KP equations is based on single runs with one reported error value per configuration. Training a neural operator is stochastic, and the reported differences between some configurations are small relative to typical run-to-run variation; for example, in Table 4.3 the errors of PODNO with 80 modes (5.0×10^-3) and FNO with 24×24 modes (4.6×10^-3) are close, and the conclusion that one method is superior would be more convincing with mean and standard deviation over several seeds. The FNO baseline is also not swept extensively, so it is unclear whether the FNO results are representative of the best achievable FNO performance for these problems. I recommend reporting repeated-seed statistics for the main comparisons and providing details of the FNO hyperparameter search.
minor comments (6)
  1. [Table A.1 and Appendix A notation] The glossary in Table A.1 uses 'Φ_k :={ϕ_k}_{k=1}^N' where the intended object appears to be the set Φ_N; the notation should be made consistent throughout the appendix.
  2. [Definitions A.1 and Theorem A.1] The paper writes 'Let Ω∈R^d' in Definition A.1 and Theorem A.1, but the domain should be a subset, i.e., 'Ω⊂R^d'. The same correction is needed in a few other places.
  3. [Appendix A equation numbering] The equations in Appendix A are numbered (1.1)–(1.3) and (2.1)–(2.2), which conflicts with the equation numbering in the main body and makes cross-referencing confusing; the appendix should use a separate numbering scheme such as (A.1), (A.2), etc.
  4. [§4.3 and Table 4.1] The text says that 30 realizations of ϵ are sampled for the NLS equation, while Table 4.1 specifies 900 training and 100 test samples; it should be clarified how the 30 values of ϵ are combined with initial-condition samples to produce the full dataset.
  5. [§4.5 and Figure 4.11] The description of the three snapshot-preparation basis types in the POD-accelerated splitting solver is clear in Algorithm 3, but the caption of Figure 4.11 is abbreviated and does not explain what 'basis type 1' refers to; please make the caption self-contained.
  6. [Throughout] There are several typographical slips, e.g., 'we will access this through numerical experiments' should be 'we will assess this through numerical experiments', and the relative errors in Tables 4.2–4.4 are reported without a formula or explicit mention that they are relative L^2 errors; adding a short note would improve clarity.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the POD basis is a data-dependent architectural choice, the universality theorem is proved for GSO from an external FNO argument, and the numerical claims rest on held-out tests against FNO.

full rationale

The paper's derivation chain is not circular. PODNO replaces the Fourier transform in the FNO kernel layer with a POD basis computed from training snapshots via SVD (Algorithm 1, Eq. 3.6); this is a model-construction choice, not a fitted parameter renamed as a prediction. The universality result (Theorem 3.1) is stated and proved for the more general GSO class, with the proof explicitly building on the external FNO universality argument of Kovachki, Lanthaler, and Mishra [27], and it does not depend on the POD basis construction. The numerical comparisons for the Darcy problem, NLS, and KP equations use held-out test data and compare against FNO baselines, so the reported accuracy and efficiency gains are not forced by construction. The paper itself flags the main theoretical limitation in Remark 3.1: the implemented L2-POD basis is not shown to satisfy the H^s-orthonormality and constant-basis assumptions (Assumptions 3.2, 3.3) required by Theorem 3.1, so the Section 5 statement that PODNO 'exhibits the same universality' is a regularity gap rather than a circular reduction. Self-citations, such as [51] for the Lie-Trotter splitting solver and [44] for spectral methods, are not load-bearing for the central claim. No equation in the paper reduces to its own input, and no prediction is equivalent by construction to a fitted quantity.

Assumptions & free parameters 4 free parameters · 7 assumptions · 0 invented entities

No new physical or mathematical entities are postulated. The GSO and PODNO are architectures, not entities with independent falsifiable handles.

free parameters (4)
  • Number of retained POD modes N = 144/576 (Darcy), 80/288 (NLS), 9/144 (KP)
    Selected per problem via the energy capture ratio rho >= 99% (eq. 3.5); directly controls truncation error, model size, and runtime.
  • Energy capture threshold rho = about 99%
    Hand-set criterion in Algorithm 1 that determines N; a data-dependent model selection rule.
  • Network hyperparameters = dv=32, L=4, learning rate 1e-3, weight decay 1e-4, batch size 20, activation GeLU
    Fixed in Table 4.1 for all experiments; the reported advantage over FNO is tied to these choices.
  • NLS potential parameters = V0=120, zeta1=zeta2=60
    Chosen by hand to create high-frequency oscillations; the POD advantage is demonstrated under this particular forcing.
assumptions (7)
  • standard math Sobolev embedding Hs(Omega) -> C(Omega) for s>d/2
    Used in Prop. A.1 and Lemma A.1 to bound sup norms of basis expansions.
  • standard math Sobolev product estimate
    Prop. A.2 used in the final error estimate of Lemma A.1.
  • standard math Universal approximation theorem for ordinary neural networks with continuous, bounded, non-constant activation
    Prop. A.3 is the core approximation step for the finite-dimensional map G_hat_N; note the activation used in experiments (GeLU) is unbounded, so the stated assumption and implementation differ.
  • ad hoc to paper The orthonormal basis contains the constant function as first basis element
    Assumption 3.3 is needed in the proof of Lemma A.1 to recover coefficients as constant functions; the POD basis is not shown to have this property.
  • ad hoc to paper Basis functions lie in Hs(Omega) with s>d/2
    Assumption 3.2 is required for Theorem 3.1, but Remark 3.1 concedes the implemented POD basis is L2-based and may not satisfy this.
  • domain assumption Ground truth solutions at 64x64 with 1000 time steps are accurate enough
    The paper uses these as labels; its own ablation in Sec. 4.6.1 shows smaller time steps change the solution, so the high-frequency components of the ground truth carry solver error.
  • domain assumption POD basis from input/output snapshots remains appropriate for hidden representations in the kernel layers
    The kernel acts on learned hidden features, not on the snapshot data; the optimality argument in Sec. 3.2 is about data approximation, and the shift is not analyzed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PODNO: Proper Orthogonal Decomposition Neural Operators." pith.science (2026). https://pith.science/paper/273BLQBU

@misc{pith2026250418513,
  author       = {Pith},
  title        = {Pith review of: PODNO: Proper Orthogonal Decomposition Neural Operators},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/273BLQBU}},
  note         = {Machine review of arXiv:2504.18513}
}
read the original abstract

In this paper, we introduce Proper Orthogonal Decomposition Neural Operators (PODNO) for solving partial differential equations (PDEs) dominated by high-frequency components. Building on the structure of Fourier Neural Operators (FNO), PODNO replaces the Fourier transform with (inverse) orthonormal transforms derived from the Proper Orthogonal Decomposition (POD) method to construct the integral kernel. Due to the optimality of POD basis, the PODNO has potential to outperform FNO in both accuracy and computational efficiency for high-frequency problems. From analysis point of view, we established the universality of a generalization of PODNO, termed as Generalized Spectral Operator (GSO). In addition, we evaluate PODNO's performance numerically on dispersive equations such as the Nonlinear Schrodinger (NLS) equation and the Kadomtsev-Petviashvili (KP) equation.

Figures

Figures reproduced from arXiv: 2504.18513 by the authors.

Figure 2.1
Figure 2.1. FNO. (a) The architecture of FNO consists of 3 parts: dimension lifting P; a sequence of Fourier layers each followed by a nonlinear activation function except the final one; and dimension reduction Q. (b) Each Fourier layer is given by F−1 N [Rl ◦ FN [v(l) ]] + Wl [v(l) ]. Dimension lifting. The dimension lifting P to map the da dimensional vector field input a to its first dv dimensional vector field hidden repres… view at source ↗
Figure 3.1
Figure 3.1. Generalized Spectral Operators (GSO). (a) The architecture of GSO comprises three main components: a dimension lifting P; a sequence of GSO layers Ll that incorporate physical parameters, and apply a nonlinear activation function σ except in the final layer; and a dimension reduction Q. (b) The GSO layers Ll = Π −1 N [Rl · ΠN [v(l) ]] +Wl [v(l) ] . (c) The POD basis for a family of KP equations, which will be explic… view at source ↗
Figure 4.1
Figure 4.1. Visualization of the Darcy problem: (Top) Four realizations of random permeability fields a generated according to (4.2); (Middle) the corre￾sponding ground truth solutions; (Bottom) the PODNO predictions using the same input fields as in the top row. The use of complex numbers in the Fourier transform makes FNO less time-efficient than PODNO, as it requires doubling the number of real parameters to accommodate comp… view at source ↗
Figures from the paper (17 more)
Figure 4.2
Figure 4.2. Figure 4.2: Test errors comparison between FNO and PODNO for the Darcy problem However, as shown in [PITH_FULL_IMAGE:figures/full_fig_p012_4_2.png]
Figure 4.3
Figure 4.3. Figure 4.3: Gaussian-type wave packet initial conditions for the NLS equa￾tion, with four randomly generated instances. The top row displays the real part, while the bottom row shows the corresponding imaginary part. The ground truth solutions (see [PITH_FULL_IMAGE:figures/full…
Figure 4.4
Figure 4.4. Figure 4.4: Ground truth solutions of the NLS equation at time T = 0.5 for the initial conditions in [PITH_FULL_IMAGE:figures/full_fig_p014_4_4.png]
Figure 4.5
Figure 4.5. Figure 4.5: PODNO predictions of the NLS equation at time T = 0.5 for the initial conditions in [PITH_FULL_IMAGE:figures/full_fig_p014_4_5.png]
Figure 4.6
Figure 4.6. Figure 4.6: Comparison of numerical errors of two methods for the NLS equation 0 1000 2000 3000 Mode Number 10 3 10 2 10 1 10 0 10 1 Abs. Diff. 0 1000 2000 3000 Mode Number 10 3 10 2 10 1 10 0 10 1 Abs. Diff. 0 1000 2000 3000 Mode Number 10 3 10 2 10 1 10 0 10 1 Abs. Diff. 0 100…
Figure 4.7
Figure 4.7. Figure 4.7: Scatter plot of the absolute difference of Fourier coefficients be￾tween the ground truth and the operator prediction for the NLS equation (y-axis) against the first 3K modes arranged in ascending order of frequency (x-axis). The instances correspond to initial condi…
Figure 4.8
Figure 4.8. Figure 4.8: Visualization of the KP equation: (Top) Four randomly gener￾ated initial conditions according to (4.7); (Middle) corresponding ground truth solutions at time T = 0.3; (Bottom) PODNO predictions at time T = 0.3. The initial condition is chosen as u0(x, y) = 8x tanh (θ…
Figure 4.9
Figure 4.9. Figure 4.9: Test errors comparison between FNO and PODNO for the KP equation. the runtime for Experiment 2 is 343.87 seconds, while Experiment 3 takes 697.14 seconds. Despite having the same number of parameters, FNO takes more than twice as long to run as PODNO. This is because…
Figure 4.10
Figure 4.10. Figure 4.10: Scatter plot of the absolute difference of Fourier coefficients between the ground truth and the operator prediction for the KP equation (y-axis) against the first 3K modes arranged in ascending order of frequency (x-axis). The top row presents the results for FNO, …
Figure 4
Figure 4. Figure 4: also presents the first 3 [PITH_FULL_IMAGE:figures/full_fig_p018_4.png]
Figure 4.11
Figure 4.11. Figure 4.11: Convergence plots for POD-accelerated Lie-Trotter time-splitting solver (left) and PODNO with 900 snapshots and 500 epochs training (right). The smallest mode number is 10 in the plot. The left side of [PITH_FULL_IMAGE:figures/full_fig_p018_4_11.png]
Figure 4.12
Figure 4.12. Figure 4.12: Solutions at time T = 0.5 obtained using the Lie-Trotter time￾splitting scheme for the NLS equation, with a fixed initial condition and con￾stant ϵ, but varying time step and mesh configurations. The top row shows results on a 32 × 32 mesh, and the bottom row on a 6…
Figure 4.13
Figure 4.13. Figure 4.13: Solutions at time T = 0.3 obtained using Exponential Time Differencing Fourth-Order Runge-Kutta (ETD4RK) for the KP equation. The top row shows results on a 32 × 32 mesh, and the bottom row on a 64 × 64 mesh. From left to right, the columns correspond to 10, 100, 50…
Figure 4
Figure 4. Figure 4: compares the solutions of the KP equation under different mesh sizes and time [PITH_FULL_IMAGE:figures/full_fig_p019_4.png]
Figure 4.14
Figure 4.14. Figure 4.14: Error comparison between different spatial discretization for the NLS equations (left) and the KP equations (right) [PITH_FULL_IMAGE:figures/full_fig_p020_4_14.png]
Figure 4.15
Figure 4.15. Figure 4.15: Error comparison between different snapshot numbers for the NLS equations (left) and the KP equations (right). 0 100 200 300 400 500 Epochs 10 2 Test errors Modes=288 Modes=144 Modes=72 0 200 400 600 800 1000 Epochs 10 3 10 2 Test errors Modes=144 Modes=36 Modes=9 …
Figure 4.16
Figure 4.16. Figure 4.16: Error comparison between different modes for NLS equations (left) and KP equations (right). For the KP equations, shown on the right, even with only 9 modes, PODNO captures 99.8% energy but still performs worse compared to the larger number of modes that capture mor…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

51 extracted references · 49 canonical work pages

  1. [1]

    Berkooz, P

    G. Berkooz, P. Holmes, and J. L. Lumley , The proper orthogonal decomposition in the analysis of turbulent flows, Annual Review of Fluid Mechanics, 25 (1993), pp. 539–575

    work page 1993

  2. [2]

    Bhattacharya, B

    K. Bhattacharya, B. Hosseini, N. B. Kovachki, and A. M. Stuart , Model reduction and neural networks for parametric PDEs, The SMAI Journal of Computational Mathematics, 7 (2021), pp. 121–157

    work page 2021

  3. [3]

    S. C. Brenner and L. R. Scott , The mathematical theory of finite element methods , Springer, 2002

    work page 2002

  4. [4]

    Q. Cao, S. Goswami, and G. E. Karniadakis , Laplace neural operator for solving differential equations , Nature Machine Intelligence, 6 (2024), pp. 631–640

    work page 2024

  5. [5]

    Cheng, C.-L

    X.-P. Cheng, C.-L. Chen, and S.-Y. Lou , Interactions among different types of nonlinear waves described by the Kadomtsev–Petviashvili equation , Wave Motion, 51 (2014), pp. 1298–1308

    work page 2014

  6. [6]

    D. N. Christodoulides, F. Lederer, and Y. Silberberg , Discretizing light behaviour in linear and nonlinear waveguide lattices, Nature, 424 (2003), pp. 817–823

    work page 2003

  7. [7]

    Cottet, P

    G.-H. Cottet, P. D. Koumoutsakos, et al. , Vortex methods: Theory and practice , vol. 313, Cambridge University Press, 2000

    work page 2000

  8. [8]

    S. M. Cox and P. C. Matthews , Exponential time differencing for stiff systems, Journal of Computational Physics, 176 (2002), pp. 430–455

    work page 2002

Show all 51 references
  1. [9]

    T. Cui, Z. W ang, and Z. Zhang , A variational neural network approach for glacier modelling with non- linear rheology, arXiv preprint arXiv:2209.02088, (2022)

    2022 arXiv

  2. [10]

    B. G. da Costa and I. S. Gomez , Information-theoretic measures for a position-dependent mass system in an infinite potential well , Physica A: Statistical Mechanics and its Applications, 541 (2020), p. 123698

    2020

  3. [11]

    N. Demo, M. Tezzele, and G. Rozza , A DeepONet multi-fidelity approach for residual learning in reduced order modeling, Advanced Modeling and Simulation in Engineering Sciences, 10 (2023), p. 12

    2023

  4. [12]

    L. C. Evans , Partial differential equations, vol. 19, American Mathematical Society, 2022

    2022

  5. [13]

    Fibich , The nonlinear Schr¨ odinger equation, vol

    G. Fibich , The nonlinear Schr¨ odinger equation, vol. 192, Springer, 2015

    2015

  6. [14]

    Galati and S

    L. Galati and S. Zheng , Nonlinear Schr¨ odinger equations for Bose-Einstein condensates, in AIP Confer- ence Proceedings, vol. 1562, American Institute of Physics, 2013, pp. 50–64

    2013

  7. [15]

    Gottlieb and S

    D. Gottlieb and S. A. Orszag , Numerical analysis of spectral methods: Theory and applications , SIAM, 1977

    1977

  8. [16]

    Grava, C

    T. Grava, C. Klein, and G. Pitton , Numerical study of the Kadomtsev–Petviashvili equation and dis- persive shock waves, Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, 474 (2018), p. 20170458

    2018

  9. [17]

    Greiner, I

    M. Greiner, I. Bloch, O. Mandel, T. W. H ¨ansch, and T. Esslinger , Bose–Einstein condensates in 1d-and 2d optical lattices , Applied Physics B, 73 (2001), pp. 769–772

    2001

  10. [18]

    Griffin, D

    A. Griffin, D. W. Snoke, and S. Stringari , Bose-Einstein condensation, Cambridge University Press, 1996

    1996

  11. [19]

    Hornik , Approximation capabilities of multilayer feedforward networks , Neural Networks, 4 (1991), pp

    K. Hornik , Approximation capabilities of multilayer feedforward networks , Neural Networks, 4 (1991), pp. 251–257

    1991

  12. [20]

    Huang, V

    G. Huang, V. A. Makarov, and M. G. Velarde , Two-dimensional solitons in Bose-Einstein condensates with a disk-shaped trap , Physical Review A, 67 (2003), p. 023604

    2003

  13. [21]

    R. S. Johnson , The classical problem of water waves: A reservoir of integrable and nearly-integrable equations, Journal of Nonlinear Mathematical Physics, 10 (2003), pp. 72–92

    2003

  14. [22]

    Jones and P

    C. Jones and P. H. Roberts , Motions in a Bose condensate. IV. Axisymmetric solitary waves , Journal of Physics A: Mathematical and General, 15 (1982), p. 2599

    1982

  15. [23]

    B. B. Kadomtsev and V. I. Petviashvili , On the stability of solitary waves in weakly dispersing media , in Doklady Akademii Nauk, vol. 192, Russian Academy of Sciences, 1970, pp. 753–756

    1970

  16. [24]

    Kato , On nonlinear Schr¨ odinger equations, in Annales de l’IHP Physique th´ eorique, vol

    T. Kato , On nonlinear Schr¨ odinger equations, in Annales de l’IHP Physique th´ eorique, vol. 46, 1987, pp. 113–129

    1987

  17. [25]

    Kato and G

    T. Kato and G. Ponce , Commutator estimates and the Euler and Navier-Stokes equations , Communica- tions on Pure and Applied Mathematics, 41 (1988), pp. 891–907

    1988

  18. [26]

    Klein, C

    C. Klein, C. Sparber, and P. Markowich , Numerical study of oscillatory regimes in the Kadomtsev– Petviashvili equation, Journal of Nonlinear Science, 17 (2007), pp. 429–470. PODNO: PROPER ORTHOGONAL DECOMPOSITION NEURAL OPERATORS 23

    2007

  19. [27]

    Kovachki, S

    N. Kovachki, S. Lanthaler, and S. Mishra , On universal approximation and error bounds for Fourier neural operators, Journal of Machine Learning Research, 22 (2021), pp. 1–76

    2021

  20. [28]

    Kovachki, Z

    N. Kovachki, Z. Li, B. Liu, K. Azizzadenesheli, K. Bhattacharya, A. Stuart, and A. Anandkumar , Neural operator: Learning maps between function spaces with applications to PDEs , Journal of Machine Learning Research, 24 (2023), pp. 1–97

    2023

  21. [29]

    Kunisch and S

    K. Kunisch and S. Volkwein , Galerkin proper orthogonal decomposition methods for parabolic problems, Numerische Mathematik, 90 (2001), pp. 117–148

    2001

  22. [30]

    R. J. LeVeque and R. J. Leveque , Numerical methods for conservation laws , vol. 214, Springer, 1992

    1992

  23. [31]

    Z. Li, N. Kovachki, K. Azizzadenesheli, B. Liu, K. Bhattacharya, A. Stuart, and A. Anandkumar , Neural operator: Graph kernel network for partial differential equations , arXiv preprint arXiv:2003.03485, (2020)

    2020 arXiv

  24. [32]

    Z. Li, N. Kovachki, K. Azizzadenesheli, B. Liu, A. Stuart, K. Bhattacharya, and A. Anandku- mar, Multipole graph neural operator for parametric partial differential equations , Advances in Neural Information Processing Systems, 33 (2020), pp. 6755–6766

    2020

  25. [33]

    Z. Li, N. B. Kovachki, K. Azizzadenesheli, K. Bhattacharya, A. Stuart, A. Anandkumar, et al. , Fourier neural operator for parametric partial differential equations, in International Conference on Learn- ing Representations, 2021

    2021

  26. [34]

    Z. Liu, H. W ang, H. Zhang, K. Bao, X. Qian, and S. Song , Render unto numerics: Orthogonal poly- nomial neural operator for PDEs with nonperiodic boundary conditions , SIAM Journal on Scientific Com- puting, 46 (2024), pp. C323–C348

    2024

  27. [35]

    Z. Liu, Y. Wu, D. Z. Huang, H. Zhang, X. Qian, and S. Song , SPFNO: Spectral operator learning for PDEs with Dirichlet and Neumann boundary conditions , arXiv preprint arXiv:2312.06980, (2023)

    2023 arXiv

  28. [36]

    L. Lu, P. Jin, G. Pang, Z. Zhang, and G. E. Karniadakis , Learning nonlinear operators via DeepONet based on the universal approximation theorem of operators, Nature Machine Intelligence, 3 (2021), pp. 218– 229

    2021

  29. [37]

    Luttinger , The effect of a magnetic field on electrons in a periodic potential , Physical Review, 84 (1951), p

    J. Luttinger , The effect of a magnetic field on electrons in a periodic potential , Physical Review, 84 (1951), p. 814

    1951

  30. [38]

    Marsiglio, The harmonic oscillator in quantum mechanics: A third way , American Journal of Physics, 77 (2009), pp

    F. Marsiglio, The harmonic oscillator in quantum mechanics: A third way , American Journal of Physics, 77 (2009), pp. 253–258

    2009

  31. [39]

    Morsch and M

    O. Morsch and M. Oberthaler , Dynamics of Bose-Einstein condensates in optical lattices , Reviews of Modern Physics, 78 (2006), pp. 179–215

    2006

  32. [40]

    N. H. Nelsen and A. M. Stuart , The random feature model for input-output maps between banach spaces, SIAM Journal on Scientific Computing, 43 (2021), pp. A3212–A3243

    2021

  33. [41]

    Quarteroni and A

    A. Quarteroni and A. V alli , Numerical approximation of partial differential equations , vol. 23, Springer Science & Business Media, 2008

    2008

  34. [42]

    Raissi, P

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

    2019

  35. [43]

    Seaman, L

    B. Seaman, L. Carr, and M. Holland , Nonlinear band structure in Bose-Einstein condensates: Nonlinear Schr¨ odinger equation with a Kronig-Penney potential, Physical Review A—Atomic, Molecular, and Optical Physics, 71 (2005), p. 033622

    2005

  36. [44]

    J. Shen, T. Tang, and L.-L. W ang , Spectral methods: Algorithms, analysis and applications , vol. 41, Springer Science & Business Media, 2011

    2011

  37. [45]

    M. A. Singer and W. H. Green , Using adaptive proper orthogonal decomposition to solve the reaction– diffusion equation, Applied Numerical Mathematics, 59 (2009), pp. 272–279

    2009

  38. [46]

    Smithies , Integral equations, no

    F. Smithies , Integral equations, no. 49, CUP Archive, 1958

    1958

  39. [47]

    Veselov and S

    A. Veselov and S. Novikov , Finite-zone, two-dimensional, potential Schr¨ odinger operators. Explicit formulas and evolution equations , in Soviet Math. Dokl, vol. 30, 1984, pp. 588–591

    1984

  40. [48]

    W ang, H

    S. W ang, H. W ang, and P. Perdikaris , Learning the solution operator of parametric partial differential equations with physics-informed DeepONets , Science Advances, 7 (2021), p. eabi8605

    2021

  41. [49]

    W ang and Z

    Z. W ang and Z. Zhang , A mesh-free method for interface problems using the deep learning approach , Journal of Computational Physics, 400 (2020), p. 108963

    2020

  42. [50]

    Yu et al

    B. Yu et al. , The deep Ritz method: A deep learning-based numerical algorithm for solving variational problems, Communications in Mathematics and Statistics, 6 (2018), pp. 1–12

    2018

  43. [51]

    Zhang and L.-L

    X. Zhang and L.-L. W ang , Low regularity estimates of the Lie-Trotter time-splitting Fourier spectral method for the logarithmic Schr¨ odinger equation, Mathematics of Computation, (2025). 24 ZILAN CHENG ∗, ZHONGJIAN WANG†, LI-LIAN WANG‡, AND MEJDI AZAIEZ§ Appendix A. Proof f...

    2025

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.