Pith. sign in

REVIEW 3 major objections 6 minor 2 cited by

Local PDE stencils become O(log K)-depth quantum convolutions inside a classical multigrid solver that stays faithful to the discrete physics.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-13 18:59 UTC pith:R77ZR7ZK

load-bearing objection Solid hybrid packaging of fixed PDE stencils into LCU–QFT local quantum convolutions inside a classical W-cycle; simulator-consistent, incremental, and scoped honestly enough to deserve referees. the 3 major comments →

arxiv 2603.24196 v2 pith:R77ZR7ZK submitted 2026-03-25 quant-ph cs.LGphysics.comp-ph

Quantum Neural Physics: Solving Partial Differential Equations on Quantum Simulators using Quantum Convolutional Neural Networks

Jucai Zhai , Muhammad Abdullah , Boyang Chen , Fazal Chaudry , Paul N. Smith , Claire E. Heaney , Yanghua Wang , Jiansheng Xiang
show 1 more author
Christopher C. Pain
This is my paper
classification quant-ph cs.LGphysics.comp-ph
keywords Quantum Neural Physicsquantum convolutional neural networkshybrid quantum-classical multigridLinear Combination of UnitariesQuantum Fourier Transformamplitude encodingPDE solverscomputational fluid dynamics
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper argues that the fixed convolutional kernels of Neural Physics—analytically prescribed stencils from finite-difference or finite-element discretisations—can be realised as quantum circuits whose depth grows only as the log of the encoded block size. Using amplitude encoding, Linear Combination of Unitaries, and the Quantum Fourier Transform, each local convolution, restriction, or prolongation becomes a shallow quantum primitive. Those primitives are then scheduled by a classical W-cycle multigrid (viewed as a U-Net) so that the global solver inherits classical multigrid convergence while the heaviest local operator applications run on a quantum engine. On noiseless simulators the hybrid method recovers accurate solutions for Poisson, diffusion, convection–diffusion, and incompressible Navier–Stokes problems, including a Kármán vortex street. The practical claim is not that a full end-to-end quantum speed-up is already available, but that a matrix-free, physics-preserving path now exists from classical stencil operators to logarithmic-depth quantum circuits, offering exponential state compression and a balanced trade-off among circuit depth, numerical robustness, and PDE structure.

Core claim

Analytically fixed local stencil operators of PDE discretisations can be mapped, via amplitude encoding plus LCU and QFT, to quantum convolutional primitives of circuit depth O(log K) for an encoded block of size K; when these primitives are embedded inside a classical W-cycle multigrid, the resulting hybrid solver produces numerically consistent solutions for Poisson, transient diffusion, convection–diffusion, and incompressible Navier–Stokes problems on noiseless simulators.

What carries the argument

The LCU–QFT quantum convolution engine: a 3 imes3 stencil is written as a weighted sum of nine translation unitaries; after a Quantum Fourier Transform those translations become diagonal phase rotations, so the whole block-encoded convolution (plus matching restriction/prolongation circuits) realises a K imes K o(K−2) imes(K−2) map at O(log K) depth and is slid across the grid inside a classical W-cycle.

Load-bearing premise

That the idealized parallel O(log K) depth for local blocks, together with classical sliding-window fallback and the omission of full state-preparation and measurement costs, is enough to claim a scalable quantum-structured operator path for large PDE systems.

What would settle it

Run the same HQC-CNNMG workflow on a real or noisy quantum device (or a full-cost resource-counting simulator that includes state preparation and measurement) for a Poisson or Navier–Stokes problem large enough that preparation/measurement dominate; if residual reduction stalls or wall-clock cost exceeds classical multigrid, the claimed scalable path fails.

Watch this falsifier — get emailed when new claim-graph text bears on it.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The manuscript introduces Quantum Neural Physics and a Hybrid Quantum-Classical CNN Multigrid Solver (HQC-CNNMG). Analytically fixed FDM/ConvFEM stencils are mapped to amplitude-encoded local quantum convolutions via LCU and QFT, with idealized parallel circuit depth O(log K) for an encoded block of size K, and embedded in a classical W-cycle multigrid (U-Net-like) architecture. Quantum restriction and prolongation are realized with lightweight Hadamard circuits. On noiseless PennyLane simulations the method is applied to Poisson, transient diffusion, convection–diffusion, and incompressible Navier–Stokes (flow past a square cylinder), reporting close agreement with classical references and stable multilevel behaviour. The authors position the contribution as a structured, matrix-free operator path rather than end-to-end quantum global inversion, and note state-preparation/measurement and NISQ limitations in the conclusion.

Significance. If the construction and numerical evidence hold, the paper offers a concrete bridge between Neural Physics (fixed convolutional stencils) and quantum structured operators, with a pragmatic hybrid multigrid workflow that preserves classical multigrid robustness while localizing quantum work to shallow LCU–QFT blocks. Strengths include analytically prescribed (untrained) kernels, an explicit O(log K) circuit template for 3×3 stencils, hybrid sliding-window design with classical fallback, and multi-physics validation through to a moderate-Re Navier–Stokes case on simulators. The work is exploratory rather than a demonstrated asymptotic speedup; its value is as a reproducible algorithmic mapping and workflow study for future FTQC or improved hybrid backends, not as a replacement for classical multigrid on current hardware.

major comments (3)
  1. [§§2.3–2.4, Abstract] §§2.3–2.4 and Abstract: The O(log K) depth claim is stated under an idealized parallel circuit model for the LCU–QFT block encoding of an already amplitude-encoded K×K block. The hybrid sliding-window scheme (§2.4) requires O(N/K²) independent calls, each needing classical-to-quantum loading of the block. The manuscript does not quantify state-preparation or measurement cost per call, nor how that cost scales relative to classical O(K²) convolution. Because the abstract and introduction still invoke exponential compression and a path to acceleration, a short resource table (qubits, depth, prep/measure model, total calls for the reported grids) is needed so the scoped claim is not read as end-to-end quantum advantage.
  2. [§2.5] §2.5: Quantum restriction maps a 2×2 block to a scalar via Hadamards and the |00⟩ amplitude; prolongation uses uniform superposition. Multigrid residual transfer and error correction require controlled accuracy. The paper does not analyze shot noise, amplitude-estimation cost, or how measurement error propagates through W-cycle residual norms and coarse-grid corrections. Without this, numerical consistency on a state-vector simulator does not yet establish that the quantum restriction/prolongation primitives remain multigrid-stable under realistic sampling.
  3. [§3] §3 (esp. 3.2–3.4): Results report relative/absolute errors against SciPy or analytical solutions and residual decay within W-cycles, but do not compare multigrid convergence factors, cycle counts, or residual histories against an identical classical W-cycle with the same smoothers, restriction/prolongation, and η, ϕ. Such a side-by-side is load-bearing for the claim of “stable multilevel behaviour” and “retaining the robustness and convergence properties of classical multigrid,” as opposed to merely matching a final solution on small grids.
minor comments (6)
  1. [Abstract / Introduction] User-facing abstract (and some intro phrasing) mentions comparisons with quantum linear solver paradigms and a balanced trade-off; the body has no dedicated comparison section or table against HHL/QLSA/VQLS on the same problems. Either add a short comparison subsection or temper the abstract to match the manuscript content.
  2. [Fig. 2, §2.4] Fig. 2 reports gate count 99 and depth 53 for K=4; state explicitly whether depth is critical-path under full parallelization or a sequential simulator count, and whether ancilla uncomputation is included.
  3. [§3.6] §3.6: Clarify which operators in the NS/SIMPLE loop use quantum convolution versus classical fallback (momentum convection/diffusion vs pressure Poisson only), and what K and number of quantum calls were used on the 256×64 grid.
  4. [§2.4] Notation: K is both encoded block size and, in places, confusable with kernel; distinguish block size from stencil support consistently.
  5. [§2.1, §3.6] CFL condition (16) and immersed-boundary σ=10^8 are free parameters; a brief sensitivity note would help reproducibility.
  6. [Throughout] Minor typos and formatting: “exponentional” (intro); inconsistent spacing in O(log K); ensure all figure panels have units/colour bars where physical fields are plotted.

Circularity Check

0 steps flagged

No significant circularity: analytically fixed FDM/ConvFEM stencils map to LCU–QFT circuits by construction and are validated against independent classical solvers, not fitted or self-forced predictions.

full rationale

The derivation chain is self-contained and non-circular. Local stencil operators (e.g., five-point Laplacian K_diff, upwind convection K_conv, ConvFEM kernels) are obtained by standard second-order central/upwind or ConvFEM discretizations of the PDEs (Eqs. 3–10, 19–20); their weights are analytically prescribed by the governing equations and grid spacing h, not trained or fitted to the reported solutions. These fixed kernels are then block-encoded via LCU of nine translation unitaries plus QFT diagonal phases (Eqs. 23–26, §2.3–2.4), yielding the claimed O(log K) circuit depth under the idealized parallel model the paper itself scopes. Multigrid restriction/prolongation and W-cycle scheduling are classical numerical methods topologically identified with U-Net layers; the hybrid sliding-window fallback is an engineering device, not a predictive claim. Numerical results (relative errors 10^{-4}–10^{-6} on Poisson/linear systems, mass/peak fidelity on convection–diffusion, qualitative Kármán street) are compared to independent classical references (SciPy spsolve, analytical Gaussian pulse). Self-citations to prior Neural Physics work supply the classical stencil-as-convolution idea but do not load-bear the quantum circuit construction or the simulator accuracy claims; no uniqueness theorem, fitted parameter renamed as prediction, or definitional equivalence of output to input appears. The asymptotic caveats (overall O(N) under classical scheduling, omitted state-prep/measurement, noiseless simulators) are already stated by the authors. Score 1 reflects only the ordinary presence of overlapping-author classical citations that are not load-bearing for the quantum results.

Axiom & Free-Parameter Ledger

4 free parameters · 5 axioms · 2 invented entities

The work rests on standard multigrid theory, standard quantum primitives (amplitude encoding, LCU block encoding, QFT diagonalization of translations), and the authors’ prior Neural Physics identification of discrete operators with fixed convolutions. Free parameters are ordinary numerical choices (block size K, smoothing counts, Δt, grids), not constants fitted to manufacture the main accuracy claims. No new physical particles or forces are postulated; ‘Quantum Neural Physics’ and HQC-CNNMG are methodological constructs. Load-bearing modeling choices are the idealized parallel depth model and hybrid classical fallback for incomplete blocks.

free parameters (4)
  • Encoded block size K (e.g. K=4 for 4×4→2×2)
    Chosen by hand for qubit packing and sliding-window design; controls qubit count and number of quantum calls O(N/K²).
  • Multigrid smoothing counts η, ϕ and W-cycle recursion depth
    Standard multigrid hyperparameters set per experiment (e.g. 6–10 W-cycles reported) rather than derived.
  • Time step Δt and grid resolutions (16×24 up to 256×64)
    Problem-setup choices constrained by CFL/stability and simulator size; not fitted to invent accuracy.
  • Immersed-boundary damping σ=1e8 (NS case)
    Ad hoc large penalty to enforce solid obstacle; standard immersed-boundary practice but free.
axioms (5)
  • domain assumption Local FDM/ConvFEM stencils exactly equal fixed convolutional kernels (Neural Physics equivalence).
    §2.1; carries classical discretization error and boundary handling into the quantum engine.
  • standard math LCU + QFT implements the weighted sum of translations as a block encoding with depth O(log K) under an idealized parallel circuit model.
    §§2.3–2.4; standard quantum linear-algebra toolkit; depth claim excludes full prep/measure accounting.
  • domain assumption Geometric multigrid W-cycle with the stated restriction/prolongation operators retains classical multigrid convergence behaviour when Ax is replaced by the quantum convolution engine.
    §2.2 and experiments; assumed rather than proved with a full multigrid convergence theorem for the hybrid operator.
  • ad hoc to paper Hybrid sliding-window classical fallback on incomplete boundary blocks does not spoil global numerical consistency.
    §2.4; engineering necessity for small K and NISQ-scale circuits.
  • domain assumption Noiseless state-vector simulation is an adequate proxy for workflow-level feasibility of the hybrid algorithm.
    §3 and Conclusion; authors acknowledge NISQ I/O and noise limits.
invented entities (2)
  • Quantum Neural Physics framework no independent evidence
    purpose: Name the mapping from analytically fixed PDE stencils to quantum convolutional primitives.
    Methodological branding of the Neural Physics → quantum circuit program; not an independent physical entity.
  • HQC-CNNMG (Hybrid Quantum-Classical CNN Multigrid Solver) no independent evidence
    purpose: Concrete hybrid architecture embedding quantum convolution/restriction/prolongation in a classical W-cycle U-Net schedule.
    Algorithmic construct validated only on simulators in this paper.

pith-pipeline@v1.1.0-grok45 · 22081 in / 3564 out tokens · 45019 ms · 2026-07-13T18:59:51.967143+00:00 · methodology

0 comments
read the original abstract

Neural Physics recasts local discretisations of partial differential equations (PDEs) as fixed convolutional operators, providing a physics-preserving alternative to data-driven surrogate modelling in scientific machine learning. However, existing realizations remain largely confined to classical AI hardware and do not directly connect to quantum structured operator design. To bridge this gap, we introduce a \emph{Quantum Neural Physics} framework and develop a Hybrid Quantum-Classical CNN Multigrid Solver (HQC-CNNMG). The proposed method maps analytically prescribed stencil operators to local quantum convolutional primitives and embeds them within a classical multilevel W-cycle architecture, combining the operator-centric view of scientific ML with the numerical rigor of multigrid solvers. Using amplitude encoding together with the Linear Combination of Unitaries (LCU) and the Quantum Fourier Transform (QFT), the resulting local quantum operators admit logarithmic-depth implementation, with circuit depth scaling as $\mathcal{O}(\log K)$ for an encoded block of size $K$ under the idealized parallel circuit model considered here. Numerical experiments on Poisson, transient diffusion, convection--diffusion, and incompressible Navier--Stokes problems demonstrate numerical consistency, stable multilevel behaviour, and workflow-level feasibility on noiseless simulators. Comparisons with representative quantum linear solver paradigms further show that the main strength of HQC-CNNMG lies in its balanced trade-off among local circuit depth, numerical robustness, and compatibility with PDE structure, rather than in fully quantum global inversion.

Figures

Figures reproduced from arXiv: 2603.24196 by Boyang Chen, Christopher C. Pain, Claire E. Heaney, Fazal Chaudry, Jiansheng Xiang, Jucai Zhai, Muhammad Abdullah, Paul N. Smith, Yanghua Wang.

Figure 1
Figure 1. Figure 1: Schematic diagram of the multigrid W-Cycle iteration. The figure illustrates the recursive process from the [PITH_FULL_IMAGE:figures/full_fig_p009_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Quantum convolution circuit for K × K → (K − 2) × (K − 2) (taking K = 4 as an example). This circuit uses 8 qubits (4 ancilla + 4 data), with a gate count of 99 and a circuit depth of 53, maintaining the circuit depth at the O(log K) level. linear mapping A is precisely decomposed into a weighted sum of 9 basic translation operators: A = X 8 k=0 ck · T(d k r ,dk c ) , (23) where ck are the kernel coefficie… view at source ↗
Figure 3
Figure 3. Figure 3: Comparison of linear system solving results. The first row displays results for a grid size of [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Results of the Poisson equation solution. The first row shows the source term, the exact solution and the [PITH_FULL_IMAGE:figures/full_fig_p017_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Comparison of diffusion equation solutions ( [PITH_FULL_IMAGE:figures/full_fig_p018_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Validation of the convection-diffusion equation ( [PITH_FULL_IMAGE:figures/full_fig_p020_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Validation of the convection-diffusion equation ( [PITH_FULL_IMAGE:figures/full_fig_p020_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Simulation results of flow past a square cylinder ( [PITH_FULL_IMAGE:figures/full_fig_p021_8.png] view at source ↗

discussion (0)

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

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Lowering LCU Circuit Width through Maximum-Weight Birkhoff-von Neumann Decomposition

    quant-ph 2026-05 unverdicted novelty 6.0

    Bottleneck and greedy largest-weight variants of Birkhoff-von Neumann decomposition reduce permutation count to O(N log(1/ε)) or ~2N for dense matrices, lowering LCU ancilla width and enabling α=1 normalization.

  2. Lowering LCU Circuit Width through Maximum-Weight Birkhoff-von Neumann Decomposition

    quant-ph 2026-05 unverdicted novelty 6.0

    A bottleneck and largest-weight greedy Birkhoff-von Neumann decomposition reduces LCU permutation terms from O(N²) to O(N log(1/ε)) or ~2N, halving ancilla qubits while setting normalization constant α=1.

Reference graph

Works this paper leans on

36 extracted references · 7 canonical work pages · cited by 1 Pith paper

  1. [1]

    Quarteroni and Alberto Valli.Numerical Approximation of Partial Differential Equa- tions

    Alfio M. Quarteroni and Alberto Valli.Numerical Approximation of Partial Differential Equa- tions. Springer Publishing Company, Incorporated, 1st ed. 1994. 2nd printing edition, 2008. ISBN 3540852670

    1994

  2. [2]

    Wm. A. Wulf and Sally A. McKee. Hitting the memory wall: implications of the obvious. SIGARCH Comput. Archit. News, 23(1):20–24, 1995. doi: 10.1145/216585.216588

    doi:10.1145/216585.216588 1995

  3. [3]

    Mueller, Wolfgang E

    Jack Dongarra, Pete Beckman, Terry Moore, Patrick Aerts, Giovanni Aloisio, Jean-Claude An- dre, David Barkai, Jean-Yves Berthou, Taisuke Boku, Bertrand Braunschweig, Franck Cap- pello, Barbara Chapman, Xuebin Chi, Alok Choudhary, Sudip Dosanjh, Thom Dunning, San- dro Fiore, Al Geist, Bill Gropp, Robert Harrison, Mark Hereld, Michael Heroux, Adolfy Hoisie,...

    work page doi:10.1177/1094342010391989 2011

  4. [4]

    Smith, Ayya Alieva, Qing Wang, Michael P

    Dmitrii Kochkov, Jamie A. Smith, Ayya Alieva, Qing Wang, Michael P. Brenner, and Stephan Hoyer. Machine learning-accelerated computational fluid dynamics.Proceedings of the National Academy of Sciences, 118(21):e2101784118, 2021. doi: 10.1073/pnas.2101784118

    doi:10.1073/pnas.2101784118 2021

  5. [5]

    Learning mesh- based simulation with graph networks

    Tobias Pfaff, Meire Fortunato, Alvaro Sanchez-Gonzalez, and Peter Battaglia. Learning mesh- based simulation with graph networks. InInternational Conference on Learning Representa- tions, 2021. URLhttps://openreview.net/forum?id=roNqYL0_XP

    2021

  6. [6]

    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

  7. [7]

    Brunton, Bernd R

    Steven L. Brunton, Bernd R. Noack, and Petros Koumoutsakos. Machine learning for fluid mechanics.Annual Review of Fluid Mechanics, 52:477–508, 2020. doi: https://doi.org/10.1146/ annurev-fluid-010719-060214

    2020

  8. [8]

    Kevrekidis, Lu Lu, Paris Perdikaris, Sifan Wang, and Liu Yang

    George Em Karniadakis, Ioannis G. Kevrekidis, Lu Lu, Paris Perdikaris, Sifan Wang, and Liu Yang. Physics-informed machine learning.Nature Reviews Physics, 3(6):422–440, June 2021. doi: 10.1038/s42254-021-00314-5. 23

    doi:10.1038/s42254-021-00314-5 2021

  9. [9]

    A TensorFlow-based new high- performance computational framework for CFD.Journal of Hydrodynamics, 32(4):735–746,

    Xi-Zeng Zhao, Tian-Yu Xu, Zhou-Teng Ye, and Wei-Jie Liu. A TensorFlow-based new high- performance computational framework for CFD.Journal of Hydrodynamics, 32(4):735–746,

  10. [10]

    doi: 10.1007/s42241-020-0050-0

    work page doi:10.1007/s42241-020-0050-0

  11. [11]

    A tensorflow simulation frame- work for scientific computing of fluid flows on tensor processing units.Computer Physics Com- munications, 274:108292, 2022

    Qing Wang, Matthias Ihme, Yi-Fan Chen, and John Anderson. A tensorflow simulation frame- work for scientific computing of fluid flows on tensor processing units.Computer Physics Com- munications, 274:108292, 2022. doi: https://doi.org/10.1016/j.cpc.2022.108292

    work page doi:10.1016/j.cpc.2022.108292 2022

  12. [12]

    Solving the discretised boltzmann transport equations using neural networks: Applications in neutron transport.arXiv preprint arXiv:2301.09991, 2023

    Toby R F Phillips, Claire E Heaney, C Boyang, Andrew G Buchan, and Christopher C Pain. Solving the discretised boltzmann transport equations using neural networks: Applications in neutron transport.arXiv preprint arXiv:2301.09991, 2023

    Pith/arXiv arXiv 2023

  13. [13]

    Phillips, Claire E

    Toby R.F. Phillips, Claire E. Heaney, Boyang Chen, Andrew G. Buchan, and Christopher C. Pain. Solving the discretised neutron diffusion equations using neural networks.International Journal for Numerical Methods in Engineering, 124(21):4659–4686, 2023. doi: https://doi.org/ 10.1002/nme.7321

    work page doi:10.1002/nme.7321 2023

  14. [14]

    Heaney, and Christopher C

    Boyang Chen, Claire E. Heaney, and Christopher C. Pain. Neural Physics: Using AI libraries to develop physics-based solvers for incompressible computational fluid dynamics.Computers & Fluids, 308:106981, 2026. doi: https://doi.org/10.1016/j.compfluid.2026.106981

    work page doi:10.1016/j.compfluid.2026.106981 2026

  15. [15]

    Bezgin, Aaron B

    Deniz A. Bezgin, Aaron B. Buhendwa, and Nikolaus A. Adams. JAX-Fluids: A fully- differentiable high-order computational fluid dynamics solver for compressible two-phase flows. Computer Physics Communications, 282:108527, 2023. doi: 10.1016/j.cpc.2022.108527

    doi:10.1016/j.cpc.2022.108527 2023

  16. [16]

    Boyang Chen, Claire E Heaney, Jefferson LMA Gomes, Omar K Matar, and Christopher C Pain. Solving the discretised multiphase flow equations with interface capturing on structured grids using machine learning libraries.Computer Methods in Applied Mechanics and Engineering, 426:116974, 2024

    2024

  17. [17]

    Solving the discretised shallow water equations using non-uniform grids and machine-learning libraries

    Amin Nadimy, Boyang Chen, Zimo Chen, Claire E Heaney, and Christopher C Pain. Solving the discretised shallow water equations using non-uniform grids and machine-learning libraries. Environmental Modelling & Software, 196:106752, 2025

    2025

  18. [18]

    A discrete element solution method embed- ded within a neural network.Powder Technology, 448:120258, 2024

    Sadjad Naderi, Boyang Chen, Tongan Yang, Jiansheng Xiang, Claire E Heaney, John-Paul Latham, Yanghua Wang, and Christopher C Pain. A discrete element solution method embed- ded within a neural network.Powder Technology, 448:120258, 2024

    2024

  19. [19]

    Implementing the discontinuous-galerkin finite element method using graph neural networks with application to diffusion equations.Neural Networks, 185:107061, 2025

    Linfeng Li, Jiansheng Xiang, Boyang Chen, Claire E Heaney, Steven Dargaville, and Christo- pher C Pain. Implementing the discontinuous-galerkin finite element method using graph neural networks with application to diffusion equations.Neural Networks, 185:107061, 2025

    2025

  20. [20]

    Quantum algorithms: an overview.npj Quantum Information, 2(1):1–8, 2016

    Ashley Montanaro. Quantum algorithms: an overview.npj Quantum Information, 2(1):1–8, 2016

    2016

  21. [21]

    High-precision quantum algorithms for partial differential equations.Quantum, 5:574, 2021

    Andrew M Childs, Jin-Peng Liu, and Aaron Ostrander. High-precision quantum algorithms for partial differential equations.Quantum, 5:574, 2021

    2021

  22. [22]

    Quantum algorithm for linear systems of equations.Physical Review Letters, 103(15):150502, 2009

    Aram W Harrow, Avinatan Hassidim, and Seth Lloyd. Quantum algorithm for linear systems of equations.Physical Review Letters, 103(15):150502, 2009

    2009

  23. [23]

    Variational quantum linear solver.Quantum, 7:1188, 2023

    Carlos Bravo-Prieto, Ryan LaRose, Marco Cerezo, Yigit Subasi, Lukasz Cincio, and Patrick J Coles. Variational quantum linear solver.Quantum, 7:1188, 2023. 24

    2023

  24. [24]

    High-order quantum algorithm for solving linear differential equations

    Dominic W Berry. High-order quantum algorithm for solving linear differential equations. Journal of Physics A: Mathematical and Theoretical, 47(10):105301, 2014

    2014

  25. [25]

    A variational eigenvalue solver on a photonic quantum processor.Nature communications, 5(1):4213, 2014

    Alberto Peruzzo, Jarrod McClean, Peter Shadbolt, Man-Hong Yung, Xiao-Qi Zhou, Peter J Love, Alán Aspuru-Guzik, and Jeremy L O’brien. A variational eigenvalue solver on a photonic quantum processor.Nature communications, 5(1):4213, 2014

    2014

  26. [26]

    Maziar Raissi, Paris Perdikaris, and George E Karniadakis. Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations.Journal of Computational physics, 378:686–707, 2019

    2019

  27. [27]

    FFT,FMM,andmultigridontheroadtoexascale: Performance challenges and opportunities.Journal of Parallel and Distributed Computing, 136: 63–74, 2020

    HudaIbeid,LukeOlson,andWilliamGropp. FFT,FMM,andmultigridontheroadtoexascale: Performance challenges and opportunities.Journal of Parallel and Distributed Computing, 136: 63–74, 2020. doi: 10.1016/j.jpdc.2019.09.014

    work page doi:10.1016/j.jpdc.2019.09.014 2020

  28. [28]

    Hervé Neau, Renaud Ansart, Cyril Baudry, Yvan Fournier, Nicolas Mérigoux, Chaï Koren, Jérome Laviéville, Nicolas Renon, and Olivier Simonin. HPC challenges and opportunities of industrial-scale reactive fluidized bed simulation using meshes of several billion cells on the route of Exascale.Powder Technology, 444:120018, 2024. doi: 10.1016/j.powtec.2024.120018

    work page doi:10.1016/j.powtec.2024.120018 2024

  29. [29]

    A fast quantum mechanical algorithm for database search

    Lov K Grover. A fast quantum mechanical algorithm for database search. InProceedings of the twenty-eighth annual ACM symposium on Theory of computing, pages 212–219, 1996

    1996

  30. [30]

    Cambridge university press, 2010

    Michael A Nielsen and Isaac L Chuang.Quantum computation and quantum information. Cambridge university press, 2010

    2010

  31. [31]

    U-Net: Convolutional networks for biomedical image segmentation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-Net: Convolutional networks for biomedical image segmentation. InInternational Conference on Medical Image Computing and Computer-Assisted Intervention, pages 234–241. Springer, 2015

    2015

  32. [32]

    Imagenet classification with deep convolutional neural networks.Advances in Neural Information Processing Systems, 25, 2012

    Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton. Imagenet classification with deep convolutional neural networks.Advances in Neural Information Processing Systems, 25, 2012

    2012

  33. [33]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. InProceedings of the IEEE Conference on Computer Vision and Pattern Recogni- tion (CVPR), June 2016

    2016

  34. [34]

    Gradient-based learning ap- plied to document recognition.Proceedings of the IEEE, 86(11):2278–2324, 2002

    Yann LeCun, Léon Bottou, Yoshua Bengio, and Patrick Haffner. Gradient-based learning ap- plied to document recognition.Proceedings of the IEEE, 86(11):2278–2324, 2002

    2002

  35. [35]

    Unet++: A nested U-Net architecture for medical image segmentation

    Zongwei Zhou, Md Mahfuzur Rahman Siddiquee, Nima Tajbakhsh, and Jianming Liang. Unet++: A nested U-Net architecture for medical image segmentation. InInternational work- shop on deep learning in medical image analysis, pages 3–11. Springer, 2018

    2018

  36. [36]

    Pen- nylane: Automatic differentiation of hybrid quantum-classical computations.arXiv preprint arXiv:1811.04968, 2018

    Ville Bergholm, Josh Izaac, Maria Schuld, Christian Gogolin, Shahnawaz Ahmed, Vishnu Ajith, M Sohaib Alam, Guillermo Alonso-Linaje, Bharath AkashNarayanan, Ali Asadi, et al. Pen- nylane: Automatic differentiation of hybrid quantum-classical computations.arXiv preprint arXiv:1811.04968, 2018. 25

    Pith/arXiv arXiv 2018