pith. sign in

arxiv: 2605.27408 · v1 · pith:QSLTORTAnew · submitted 2026-05-12 · 🪐 quant-ph · cs.LG· cs.NA· math.NA

Neural Quantum Spectral Operator Learning for Solving Partial Differential Equations

Chanyoung Kim , Myeonghwan Seong , Yujin Kim , Daniel K. Park , Youngjoon Hong This is my paper

Pith reviewed 2026-06-30 22:33 UTC · model grok-4.3

classification 🪐 quant-ph cs.LGcs.NAmath.NA
keywords quantum operator learningvariational quantum linear solverLegendre-Galerkin methodparametric PDEsunsupervised learningneural embeddinghybrid quantum-classicalsign ambiguity resolution
0
0 comments X

The pith

NVQLS resolves sign ambiguity in variational quantum linear solvers and adds neural embeddings to enable unsupervised operator learning for parametric PDEs.

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

The paper introduces NVQLS as a hybrid quantum-classical framework that learns solution operators for families of PDEs without needing paired input-output datasets from classical solvers. It builds on the Legendre-Galerkin weak form to set up variational quantum linear systems, then fixes the sign ambiguity that otherwise produces wrong solution signs during energy minimization. A neural embedding layer encodes varying forcings and coefficients directly into parameterized quantum circuits, allowing the same trained model to handle multiple inputs at once. The resulting method reports higher accuracy than a classical baseline while claiming theoretical complexity gains when state preparation is efficient.

Core claim

NVQLS is the first hybrid quantum-classical operator learning method that uses the Legendre-Galerkin weak formulation together with a sign-resolved variational quantum linear solver and a neural embedding scheme; the combination lets the framework map varying PDE inputs into quantum circuit parameters and produce accurate surrogate solutions for 1D and 2D parametric problems under diverse boundary conditions.

What carries the argument

The neural embedding scheme, which maps forcing functions and PDE coefficients into parameterized quantum circuit representations while the sign-resolved VQLS supplies the linear solve step inside the Legendre-Galerkin formulation.

If this is right

  • A single trained NVQLS model can process multiple different forcings and coefficients at inference time without retraining.
  • The method offers a route to unsupervised quantum-enhanced operator learning that avoids generating large classical training datasets.
  • Theoretical complexity advantages appear once efficient state-preparation oracles are available for the embedded inputs.
  • The framework extends to both 1D and 2D problems with various boundary conditions while staying within the variational quantum linear solver setting.

Where Pith is reading between the lines

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

  • If the neural embedding generalizes beyond the tested 1D/2D cases, the same architecture could be applied to time-dependent or nonlinear parametric PDEs by extending the weak-form discretization.
  • The sign-resolution technique may transfer to other variational quantum algorithms that minimize energies over signed solution spaces.
  • Scaling to higher-dimensional PDEs would require checking whether the quantum circuit depth remains practical under realistic hardware noise.

Load-bearing premise

The neural embedding successfully encodes different forcings and coefficients into quantum circuits without adding errors large enough to break the operator learning, and the sign fix in VQLS actually prevents incorrect solution representations for the PDEs tested.

What would settle it

Run NVQLS and the classical baseline on the same family of 2D parametric Poisson equations, measure L2 error on held-out coefficient/forcing pairs, and check whether the quantum method's error remains lower when the sign ambiguity is deliberately left unresolved.

Figures

Figures reproduced from arXiv: 2605.27408 by Chanyoung Kim, Daniel K. Park, Myeonghwan Seong, Youngjoon Hong, Yujin Kim.

Figure 1
Figure 1. Figure 1: Training procedure of the NVQLS framework. An angle network encodes a batch of PDE instances into quantum circuit parameters θ, while a loss function based on quantum state overlap is utilized to minimize weak-form residuals. Each component of the loss function is evaluated through quantum subroutines with the measurement grouping. where a spatial derivative operator F can be either linear or nonlinear, an… view at source ↗
Figure 2
Figure 2. Figure 2: Illustration of the reflection error inherent in standard VQLS and the effect of phase correction. Rows from top to bottom display (a–c) the initial VQLS predictions exhibiting reflection errors, and (d–f) the corresponding solutions after applying the phase correction. Columns from left to right correspond to reflections along the x-axis, the y-axis, and both axes, respectively. 2.2 VQLS As a heuristic al… view at source ↗
Figure 3
Figure 3. Figure 3: Ablation study on the effect of Neural Embedding (NE). Plots (a–d) show convergence histories for training/test losses and relative L2 errors. We compare NVQLS with NE (7,780 parameters, depth 10) using D = 50 against two quantum baselines (1,200 parameters, depth 100) evaluated across D ∈ {20, 50, 100}. Here, D denotes the number of training instances. The results highlight that neural embedding enables e… view at source ↗
Figure 4
Figure 4. Figure 4: Numerical results for 1D steady-state elliptic PDEs. Top row: (a–c) Predicted solution uˆ versus exact solution u. Bottom row: (d–f) Absolute error |uˆ − u|. Columns from left to right correspond to the reaction–diffusion (ϵ = 0.1, Dirichlet BC), Helmholtz (k = 4.7, Neumann BC), and convection–diffusion (ϵ = 0.1, Dirichlet BC) equations, respectively. is in O [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Representative solution profiles for the wave equation. Row 1: Spatio-temporal profiles of (a) the ground truth, (b) the PI-DON prediction, (c) the NVQLS prediction, and (d) the absolute error of the NVQLS prediction. Row 2: (e–h) Snapshots comparing the absolute errors of each model at t = 0.4, 0.8, 1.2, and 1.6. Notably, across 200 test samples, even the worst-case prediction of NVQLS demonstrates higher… view at source ↗
Figure 6
Figure 6. Figure 6: Numerical examples of operator learning with joint parameter and forcing inputs for the two-dimensional Helmholtz equation (Dirichlet BC). Row 1 (Case 1): (a) Input pairs f1 and k 2 = 4.032 for the angle network, (b) ground truth u1, (c) NVQLS prediction uˆ1, and (d) absolute error |uˆ1−u1| on interior nodal points. Row 2 and Row 3 show the corresponding results for Case 2 (f2 and k 2 = 4.030) and Case 3 (… view at source ↗
Figure 7
Figure 7. Figure 7: Empirical analysis of the required Pauli measurements. Comparison between NVQLS with measurement grouping (blue), without grouping (green), and the classical VQLS baseline (red). Scaling with respect to (a)-(b) the number of basis functions N for 1D and 2D cases, and (c)-(d) the dimension d for N = 5 and N = 9. C Detailed Complexity Analysis We provide a detailed complexity analysis for the results discuss… view at source ↗
Figure 8
Figure 8. Figure 8: Empirical analysis of the number of required Pauli measurements in NVQLS framework with grouping of commuting measurement operators compared with VQLS of the Helmholtz equation (k 2 = 10.6) and convection diffusion equation (ϵ = 0.05). Top: number of Pauli terms of A for increasing resolution (as a function of the number of basis functions N): (a) one-dimensional case, (b) two-dimensional case. Bottom: num… view at source ↗
Figure 9
Figure 9. Figure 9: Ansätze used in our study. (a) Strongly Entangling Layer, (b) Hardware efficient RY ansätz [PITH_FULL_IMAGE:figures/full_fig_p019_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Efficiency analysis of neural embedding under comparable parameter budgets. The plots illustrate histories over training epochs for (a) training cost, (b) testing cost, (c) relative L 2 training error, and (d) relative L 2 testing error. We compare NVQLS (1,148 pa￾rameters, depth 2) against the quantum baselines (1,200 parameters, depth 100). The results demonstrate that neural embedding provides enhanced… view at source ↗
Figure 11
Figure 11. Figure 11: (a) Representative OOD prediction of the solution u(x) for a forcing function in Eq. 31. (b) Absolute error |uˆ(x) − u(x)| for the same example. (c) Train and test relative L2 errors on the shallow Ry construction. (d) Empirical cumulative distribution functions (ECDFs) of relative L2 and L∞ errors on 10,000 OOD forcing functions. We first analyze the effect of truncating the Pauli decomposition of A by r… view at source ↗
Figure 12
Figure 12. Figure 12: Impact of the Pauli truncation threshold on the numerical properties of the truncated operator and the resulting solution accuracy for the one-dimensional Helmholtz equation with the homogeneous Dirichlet boundary condition and a wave number k 2 = 4. Top: (a) Relative Frobenius-norm error between A and A˜, condition number of A˜, and number of retained Pauli terms as functions of the truncation threshold.… view at source ↗
Figure 13
Figure 13. Figure 13: Scaling behavior of the number of Pauli terms in the truncated operator A˜ and the corresponding measurement cost for loss evaluation in the one-dimensional Helmholtz equation with Dirichlet boundary conditions. (a) Number of retained Pauli terms. (b,c) Number of Pauli measurements required for evaluating β and γ. accuracy level, the number of retained Pauli terms in A˜ scales linearly with the system siz… view at source ↗
Figure 14
Figure 14. Figure 14: Numerical results for the one-dimensional Helmholtz equation with the homo￾geneous Dirichlet boundary condition and a wave number k 2 = 29.4. Top: (a) example of the predicted solution uˆ compared to the exact solution u, (b) absolute error between uˆ and u. Bottom: (c) batch-wise training and test losses, (d) batch-wise relative L 2 and L∞ errors over epochs. 1D Convection–diffusion Equation. We now focu… view at source ↗
Figure 15
Figure 15. Figure 15: Visualization of the training and testing error curves for the proposed NVQLS and the PI-DON baseline on the 1D wave equation. 1D Wave Equation. We trained NVQLS on the 1D wave equation that is time-dependent and hyperbolic, defined as utt − uxx = f(x, t), (x, t) ∈ [0, 1] × [0, T], (40) with the homogeneous Dirichlet boundary conditions and zero initial conditions (i.e., u(x, 0) = ut(x, 0) = 0). Here, we … view at source ↗
Figure 16
Figure 16. Figure 16: Numerical results for two-dimensional PDEs. Columns from left to right display the external forcing f, the exact ground truth solution u, the NVQLS prediction uˆ, and the pointwise absolute error |uˆ − u|, respectively. Rows from top to bottom correspond to the evaluated physical systems: (a)-(d) the reaction–diffusion equation with ϵ = 0.1 (Dirichlet BC), (e)-(h)the Helmholtz equation with k 2 = 8.9 (Neu… view at source ↗
Figure 17
Figure 17. Figure 17: Numerical examples of operator learning with joint parameter and forcing inputs for the Helmholtz equation under Dirichlet boundary conditions. Top: input pairs for the angle network: (a) case 1: a forcing function f1 and a wave number k 2 = 4.457, (b) case 2: f2 and k 2 = 4.310, (c) case 3: f3 and k 2 = 4.408. Middle: exact solutions ui and predicted solutions uˆi for each case: (d) case 1, (e) case 2, (… view at source ↗
read the original abstract

Partial differential equations (PDEs) are central to modeling physical and engineering systems, but repeatedly solving parametric PDEs remains computationally expensive. Operator learning enables fast surrogate inference, yet typically requires large input-output paired datasets generated by costly high-fidelity PDE solvers. Unsupervised operator learning frameworks alleviate data dependency but remain hindered by computational bottlenecks. To address this, we propose Neural Variational Quantum Linear Solver (NVQLS), the first hybrid quantum-classical operator learning framework leveraging the Legendre--Galerkin weak formulation. We critically resolve the sign ambiguity in VQLS energy minimization, preventing erroneous solution representations. Additionally, we introduce a neural embedding, a novel encoding scheme to map varying forcings and PDE coefficients into parameterized quantum circuit representations. These structural innovations provide theoretical computational complexity advantages under efficient state preparation schemes, while achieving superior accuracy compared to a representative classical baseline. Validations on 1D and 2D parametric PDEs under diverse boundary conditions demonstrate NVQLS's capability to simultaneously process varying inputs, offering a scalable unsupervised approach to quantum-enhanced operator learning.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 2 minor

Summary. The manuscript proposes Neural Variational Quantum Linear Solver (NVQLS), presented as the first hybrid quantum-classical operator learning framework for parametric PDEs. It combines the Legendre-Galerkin weak formulation with a variational quantum linear solver (VQLS), resolves sign ambiguity in the VQLS energy minimization, introduces a neural embedding to encode varying forcings and PDE coefficients into parameterized quantum circuits, and claims both superior accuracy over a classical baseline and theoretical computational complexity advantages (conditioned on efficient state preparation) demonstrated on 1D and 2D test problems under diverse boundary conditions.

Significance. If the accuracy results and complexity claims hold after verification, the work would introduce a novel unsupervised quantum-classical route to operator learning that reduces reliance on paired high-fidelity data. The sign-ambiguity resolution and neural embedding constitute concrete technical contributions. However, the significance is limited by the absence of explicit quantitative validation details and by the central dependence on unverified efficient state preparation for the embedding, which is a known hard problem in the worst case.

major comments (2)
  1. [Abstract] Abstract: the claim of 'theoretical computational complexity advantages under efficient state preparation schemes' is load-bearing for the paper's positioning as a quantum-enhanced method, yet the neural embedding is not accompanied by any cost analysis, circuit-depth bounds, or verification that the output states admit sub-exponential preparation; general state preparation is #P-hard, so the advantage is conditional and currently unsupported.
  2. [Validation sections] Validation sections (referenced in abstract as 1D/2D experiments): the abstract asserts 'superior accuracy' and 'demonstrate NVQLS's capability' but supplies no error metrics, convergence rates, comparison tables, or statistical details; without these, the accuracy claim cannot be evaluated and is not yet load-bearing evidence.
minor comments (2)
  1. Notation for the neural embedding and the resolved sign-ambiguity term should be introduced with explicit equations rather than descriptive prose only.
  2. [Abstract] The classical baseline used for the accuracy comparison is not named in the abstract; it should be identified early.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for their detailed review and constructive comments. We address each major comment below, providing clarifications and indicating revisions where appropriate.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim of 'theoretical computational complexity advantages under efficient state preparation schemes' is load-bearing for the paper's positioning as a quantum-enhanced method, yet the neural embedding is not accompanied by any cost analysis, circuit-depth bounds, or verification that the output states admit sub-exponential preparation; general state preparation is #P-hard, so the advantage is conditional and currently unsupported.

    Authors: The manuscript explicitly conditions the complexity advantages on efficient state preparation schemes, as stated in the abstract and throughout the text. The neural embedding is presented as a structural innovation that maps varying inputs to parameterized circuits in a manner compatible with such schemes. We agree that general state preparation is hard and do not claim unconditional advantage; the contribution is the hybrid framework under the stated assumption, consistent with standard practice in quantum algorithm papers. We will expand the discussion section with explicit assumptions and references to state-preparation literature in the revision. revision: partial

  2. Referee: [Validation sections] Validation sections (referenced in abstract as 1D/2D experiments): the abstract asserts 'superior accuracy' and 'demonstrate NVQLS's capability' but supplies no error metrics, convergence rates, comparison tables, or statistical details; without these, the accuracy claim cannot be evaluated and is not yet load-bearing evidence.

    Authors: The manuscript body contains validation sections reporting quantitative results for the 1D and 2D parametric PDE test cases, including direct comparisons against a classical baseline that demonstrate superior accuracy under the tested boundary conditions. The abstract provides only a high-level summary, as is conventional. To strengthen visibility of the evidence, we will insert a consolidated error-metrics table and convergence details into the main text during revision. revision: yes

standing simulated objections not resolved
  • The central dependence on efficient state preparation for the neural embedding, which is a known hard problem in the worst case and remains unverified in general.

Circularity Check

0 steps flagged

No significant circularity; framework derivation is self-contained

full rationale

The paper proposes NVQLS as a novel hybrid framework combining Legendre-Galerkin weak form, sign ambiguity resolution in VQLS, and a neural embedding scheme. Claims of theoretical complexity advantages are explicitly conditioned on external 'efficient state preparation schemes' rather than derived by construction from the method itself. No self-definitional reductions, fitted inputs renamed as predictions, or load-bearing self-citations appear in the abstract or described claims. The derivation introduces new components without reducing the central results to reparameterizations of the inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 2 invented entities

Based solely on abstract; no specific numerical free parameters are mentioned. The framework rests on domain assumptions about quantum linear solvers and spectral methods.

axioms (2)
  • domain assumption Legendre-Galerkin weak formulation applies to the target parametric PDEs
    Central to the proposed operator learning approach
  • domain assumption VQLS energy minimization can be made reliable by resolving sign ambiguity
    Explicitly addressed as a critical resolution step
invented entities (2)
  • NVQLS framework no independent evidence
    purpose: Hybrid quantum-classical operator learning for PDEs
    New proposed method combining VQLS with neural embedding
  • neural embedding no independent evidence
    purpose: Encoding scheme to map inputs to parameterized quantum circuits
    Novel encoding introduced in the paper

pith-pipeline@v0.9.1-grok · 5733 in / 1241 out tokens · 23008 ms · 2026-06-30T22:33:01.572916+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

52 extracted references · 8 canonical work pages · 4 internal anchors

  1. [1]

    Deep learning using rectified linear units (relu), 2019

    Abien Fred Agarap. Deep learning using rectified linear units (relu), 2019

    2019

  2. [2]

    Performance study of variational quantum algorithms for solving the poisson equation on a quantum computer

    Mazen Ali and Matthias Kabel. Performance study of variational quantum algorithms for solving the poisson equation on a quantum computer. Physical Review Applied , 20 (1):014054, 2023

    2023

  3. [3]

    Ville Bergholm, Josh Izaac, Maria Schuld, Christian Gogolin, M. Sohaib Alam, Shahnawaz Ahmed, Juan Miguel Arrazola, Carsten Blank, Alain Delgado, Soran Jahangiri, Keri McK- iernan, Johannes Jakob Meyer, Zeyue Niu, Antal Száva, and Nathan Killoran. Penny- lane: Automatic differentiation of hybrid quantum-classical computations. arXiv preprint arXiv:1811.04...

    work page internal anchor Pith review Pith/arXiv arXiv 2020

  4. [4]

    Jax: composable transformations of python+ numpy programs

    James Bradbury, Roy Frostig, Peter Hawkins, Matthew James Johnson, Chris Leary, Dougal Maclaurin, George Necula, Adam Paszke, Jake VanderPlas, Skye Wanderman-Milne, et al. Jax: composable transformations of python+ numpy programs. 2018

    2018

  5. [5]

    Variational quantum linear solver

    Carlos Bravo-Prieto, Ryan LaRose, Marco Cerezo, Yigit Subasi, Lukasz Cincio, and Patrick J Coles. Variational quantum linear solver. arXiv preprint arXiv:1909.05820 , 2019

    work page arXiv 1909

  6. [6]

    Variational quantum linear solver

    Carlos Bravo-Prieto, Ryan LaRose, Marco Cerezo, Yigit Subasi, Lukasz Cincio, and Patrick J Coles. Variational quantum linear solver. Quantum, 7:1188, 2023

    2023

  7. [7]

    Burden and J

    Richard L. Burden and J. Douglas Faires. Numerical Analysis . The Prindle, Weber and Schmidt Series in Mathematics. PWS-Kent Publishing Company, Boston, fourth edition, 1989

    1989

  8. [8]

    Childs and Jin-Peng Liu

    Andrew M. Childs and Jin-Peng Liu. Quantum spectral methods for differential equations. Communications in Mathematical Physics , 375 (2):1427–1457, 2020

    2020

  9. [9]

    Physics-informed deep inverse operator networks for solving pde inverse problems, 2025

    Sung Woong Cho and Hwijae Son. Physics-informed deep inverse operator networks for solving pde inverse problems, 2025

    2025

  10. [10]

    Spectral operator learning for parametric pdes without data reliance

    Junho Choi, Taehyun Yun, Namjung Kim, and Youngjoon Hong. Spectral operator learning for parametric pdes without data reliance. Computer Methods in Applied Mechanics and Engineering, 420:116678, 2024

    2024

  11. [11]

    Efficient variational quantum linear solver for structured sparse matrices

    Abeynaya Gnanasekaran and Amit Surana. Efficient variational quantum linear solver for structured sparse matrices. In 2024 IEEE International Conference on Quantum Computing and Engineering (QCE), volume 1, pages 199–210. IEEE, 2024

    2024

  12. [12]

    Physics-informed neural networks with domain decomposition for the incompressible navier–stokes equations

    Linyan Gu, Shanlin Qin, Lei Xu, and Ronglian Chen. Physics-informed neural networks with domain decomposition for the incompressible navier–stokes equations. Physics of Fluids , 36(2):021914, 02 2024

    2024

  13. [13]

    Ib-uq: Information bottleneck based uncertainty quantification for neural function regression and neural operator learning, 2023

    Ling Guo, Hao Wu, Wenwen Zhou, Yan Wang, and Tao Zhou. Ib-uq: Information bottleneck based uncertainty quantification for neural function regression and neural operator learning, 2023

    2023

  14. [14]

    Multiwavelet-based operator learning for dif- ferential equations

    Gaurav Gupta, Xiongye Xiao, and Paul Bogdan. Multiwavelet-based operator learning for dif- ferential equations. In Proceedings of the 35th International Conference on Neural Information Processing Systems, NIPS ’21, Red Hook, NY, USA, 2021. Curran Associates Inc

    2021

  15. [15]

    Harrow, A vinatan Hassidim, and Seth Lloyd

    Aram W. Harrow, A vinatan Hassidim, and Seth Lloyd. Quantum algorithm for linear systems of equations. Phys. Rev. Lett. , 103:150502, Oct 2009

    2009

  16. [16]

    Gaussian Error Linear Units (GELUs)

    Dan Hendrycks and Kevin Gimpel. Gaussian error linear units (gelus). arXiv preprint arXiv:1606.08415, 2016

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  17. [17]

    Hosaka, K

    Aruto Hosaka, Koichi Yanagisawa, Shota Koshikawa, Isamu Kudo, Xiafukaiti Alifu, and Tsuyoshi Yoshida. Preconditioning for a variational quantum linear solver. arXiv preprint arXiv:2312.15657, 2023

    work page arXiv 2023

  18. [18]

    Tackling the curse of dimensionality with physics-informed neural networks

    Zheyuan Hu, Khemraj Shukla, George Em Karniadakis, and Kenji Kawaguchi. Tackling the curse of dimensionality with physics-informed neural networks. Neural Networks , 176:106369, 2024

    2024

  19. [19]

    Quantum fourier networks for solving parametric pdes

    Nishant Jain, Jonas Landman, Natansh Mathur, and Iordanis Kerenidis. Quantum fourier networks for solving parametric pdes. Quantum Science and Technology , 9 (3):035026, 2024. 10

    2024

  20. [20]

    Kharazmi, Z

    E. Kharazmi, Z. Zhang, and G. E. Karniadakis. Variational physics-informed neural networks for solving partial differential equations, 2019

    2019

  21. [21]

    Adam: A Method for Stochastic Optimization

    Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 , 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  22. [22]

    Abueidda, Nahil A

    Seid Koric, Asha Viswantah, Diab W. Abueidda, Nahil A. Sobh, and Kamran Khan. Deep learning operator network for plastic deformation with variable loads and material properties. Engineering with Computers , 40 (2):917–929, 2024

    2024

  23. [23]

    Neural operator: learning maps between function spaces with applications to pdes

    Nikola Kovachki, Zongyi Li, Burigede Liu, Kamyar Azizzadenesheli, Kaushik Bhattacharya, Andrew Stuart, and Anima Anandkumar. Neural operator: learning maps between function spaces with applications to pdes. J. Mach. Learn. Res. , 24 (1), January 2023

    2023

  24. [24]

    Paine, and Vincent E

    Oleksandr Kyriienko, Annie E. Paine, and Vincent E. Elfving. Solving nonlinear differential equations with differentiable quantum circuits. Phys. Rev. A , 103:052416, May 2021

    2021

  25. [25]

    Spectral coefficient learning via operator networks for inverse problems of parametric partial differential equations

    Myeong-Su Lee, Taehyun Yun, Youngjoon Hong, and Namjung Kim. Spectral coefficient learning via operator networks for inverse problems of parametric partial differential equations. Engineering Applications of Artificial Intelligence , 166:113665, 2026

    2026

  26. [26]

    Fourier Neural Operator for Parametric Partial Differential Equations

    Zongyi Li, Nikola B. Kovachki, Kamyar Azizzadenesheli, Burigede Liu, Kaushik Bhattacharya, Andrew M. Stuart, and Anima Anandkumar. Fourier neural operator for parametric partial differential equations. arXiv preprint arXiv:2010.08895 , 2020

    work page internal anchor Pith review Pith/arXiv arXiv 2010

  27. [27]

    Kovachki, David Jin, Haoxuan Chen, Burigede Liu, Kam- yar Azizzadenesheli, and Anima Anandkumar

    Zongyi Li, Hongkai Zheng, Nikola B. Kovachki, David Jin, Haoxuan Chen, Burigede Liu, Kam- yar Azizzadenesheli, and Anima Anandkumar. Physics-informed neural operator for learning partial differential equations. arXiv preprint arXiv:2111.03794 , 2021

    work page arXiv 2021

  28. [28]

    Liu and Jorge Nocedal

    Dong C. Liu and Jorge Nocedal. On the limited memory bfgs method for large scale optimiza- tion. Mathematical Programming, 45:503–528, 1989

    1989

  29. [29]

    Variational quantum algorithm for the poisson equation

    Hai-Ling Liu, Yu-Sen Wu, Lin-Chun Wan, Shi-Jie Pan, Su-Juan Qin, Fei Gao, and Qiao- Yan Wen. Variational quantum algorithm for the poisson equation. Physical Review A , 104(2):022418, 2021

    2021

  30. [30]

    Application of a variational hybrid quantum-classical algorithm to heat conduction equation and analysis of time complexity

    YY Liu, Zhen Chen, Chang Shu, Siou Chye Chew, Boo Cheong Khoo, Xiang Zhao, and YD Cui. Application of a variational hybrid quantum-classical algorithm to heat conduction equation and analysis of time complexity. Physics of Fluids , 34 (11), 2022

    2022

  31. [31]

    Learning nonlinear operators via deeponet based on the universal approximation theorem of operators

    Lu Lu, Pengzhan Jin, Guofei Pang, Zhongqiang Zhang, and George Em Karniadakis. Learning nonlinear operators via deeponet based on the universal approximation theorem of operators. Nature machine intelligence , 3 (3):218–229, 2021

    2021

  32. [32]

    Menon and Ameya D

    Sidharth S. Menon and Ameya D. Jagtap. Anant-net: Breaking the curse of dimensionality with scalable and interpretable neural surrogate for high-dimensional pdes. Computer Methods in Applied Mechanics and Engineering , 447:118403, 2025

    2025

  33. [33]

    Mauro E. S. Morales, Lirandë Pira, Philipp Schleich, Kelvin Koor, Pedro C. S. Costa, Dong An, Alán Aspuru-Guzik, Lin Lin, Patrick Rebentrost, and Dominic W. Berry. Quantum linear system solvers: A survey of algorithms and applications, 2025

    2025

  34. [34]

    Deep-learning-based quantum algorithms for solving nonlinear partial differential equations

    Lukas Mouton, Florentin Reiter, Ying Chen, and Patrick Rebentrost. Deep-learning-based quantum algorithms for solving nonlinear partial differential equations. Phys. Rev. A , 110:022612, Aug 2024

    2024

  35. [35]

    Variational quantum linear solver with a dynamic ansatz

    Hrushikesh Patil, Yulun Wang, and Predrag S Krstić. Variational quantum linear solver with a dynamic ansatz. Physical Review A , 105 (1):012423, 2022

    2022

  36. [36]

    A comparison of various classical optimizers for a variational quantum linear solver

    Aidan Pellow-Jarman, Ilya Sinayskiy, Anban Pillay, and Francesco Petruccione. A comparison of various classical optimizers for a variational quantum linear solver. Quantum Information Processing, 20 (6):202, 2021

    2021

  37. [37]

    Near term algorithms for linear systems of equations

    Aidan Pellow-Jarman, Ilya Sinayskiy, Anban Pillay, and Francesco Petruccione. Near term algorithms for linear systems of equations. Quantum Information Processing , 22 (6):258, 2023

    2023

  38. [38]

    Rudy, Steven L

    Samuel H. Rudy, Steven L. Brunton, Joshua L. Proctor, and J. Nathan Kutz. Data-driven discovery of partial differential equations. Science Advances, 3 (4):e1602614, 2017. 11

    2017

  39. [39]

    Hybrid quantum physics-informed neural networks for simulating com- putational fluid dynamics in complex shapes

    Alexandr Sedykh, Maninadh Podapaka, Asel Sagingalieva, Karan Pinto, Markus Pflitsch, and Alexey Melnikov. Hybrid quantum physics-informed neural networks for simulating com- putational fluid dynamics in complex shapes. Machine Learning: Science and Technology , 5(2):025045, may 2024

    2024

  40. [40]

    Spectral Methods: Algorithms, Analysis and Applica- tions, volume 41 of Springer Series in Computational Mathematics

    Jie Shen, Tao Tang, and Li-Lian Wang. Spectral Methods: Algorithms, Analysis and Applica- tions, volume 41 of Springer Series in Computational Mathematics . Springer Berlin, Heidelberg, 1 edition, 2011

    2011

  41. [41]

    Variational quantum framework for partial differ- ential equation constrained optimization

    Amit Surana and Abeynaya Gnanasekaran. Variational quantum framework for partial differ- ential equation constrained optimization. ACM Transactions on Quantum Computing , 2024

    2024

  42. [42]

    Quantum physics-informed neural networks

    Corey Trahan, Mark Loveland, and Samuel Dent. Quantum physics-informed neural networks. Entropy, 26 (8), 2024

    2024

  43. [43]

    A variational quan- tum linear solver application to discrete finite-element methods

    Corey Jason Trahan, Mark Loveland, Noah Davis, and Elizabeth Ellison. A variational quan- tum linear solver application to discrete finite-element methods. Entropy, 25 (4), 2023

    2023

  44. [44]

    Trefethen

    Lloyd N. Trefethen. Spectral Methods in MATLAB . Society for Industrial and Applied Math- ematics, 2000

    2000

  45. [45]

    An em- pirical analysis on the effectiveness of the variational quantum linear solver

    Gloria Turati, Alessia Marruzzo, Maurizio Ferrari Dacrema, and Paolo Cremonesi. An em- pirical analysis on the effectiveness of the variational quantum linear solver. arXiv preprint arXiv:2409.06339, 2024

    work page arXiv 2024

  46. [46]

    Learning the solution operator of para- metric partial differential equations with physics-informed deeponets

    Sifan Wang, Hanwen Wang, and Paris Perdikaris. Learning the solution operator of para- metric partial differential equations with physics-informed deeponets. Science Advances , 7(40):eabi8605, 2021

    2021

  47. [47]

    Understanding the difficulty of solving cauchy problems with pinns

    Tao Wang, Bo Zhao, Sicun Gao, and Rose Yu. Understanding the difficulty of solving cauchy problems with pinns. In Alessandro Abate, Mark Cannon, Kostas Margellos, and Antonis Papachristodoulou, editors, Proceedings of the 6th Annual Learning for Dynamics & Control Conference, volume 242 of Proceedings of Machine Learning Research, pages 453–465. PMLR, 15–1...

    2024

  48. [48]

    Solving high- dimensional partial differential equations using tensor neural network and a posteriori error estimators

    Yifan Wang, Zhongshuo Lin, Yangfei Liao, Haochen Liu, and Hehu Xie. Solving high- dimensional partial differential equations using tensor neural network and a posteriori error estimators. J. Sci. Comput. , 101 (3), November 2024

    2024

  49. [49]

    Quantum deeponet: Neural operators accelerated by quantum computing

    Pengpeng Xiao, Muqing Zheng, Anran Jiao, Xiu Yang, and Lu Lu. Quantum deeponet: Neural operators accelerated by quantum computing. Quantum, 9:1761, 2025

    2025

  50. [50]

    Analyses of internal structures and defects in materials using physics-informed neural networks

    Enrui Zhang, Ming Dao, George Em Karniadakis, and Subra Suresh. Analyses of internal structures and defects in materials using physics-informed neural networks. Science Advances, 8(7):eabk0644, 2022

    2022

  51. [51]

    A graph-structured, physics-informed deeponet neural network for complex structural analysis

    Guangya Zhang, Tie Xu, Jinli Xu, and Hu Wang. A graph-structured, physics-informed deeponet neural network for complex structural analysis. Machine Learning and Knowledge Extraction, 7 (4), 2025

    2025

  52. [52]

    Psaros, and George E

    Zongren Zou, Xuhui Meng, Apostolos F. Psaros, and George E. Karniadakis. Neuraluq: A com- prehensive library for uncertainty quantification in neural differential equations and operators. SIAM Review , 66 (1):161–190, 2024. 12 A Nomenclature Table 2: Notations and Descriptions Notation Description n Number of qubits N Number of basis functions d Dimension...

    2024