pith. sign in

arxiv: 2506.14792 · v1 · submitted 2025-05-29 · 🧮 math.NA · astro-ph.SR· cs.NA· physics.ao-ph· physics.flu-dyn

Fast automated adjoints for spectral PDE solvers

Calum S. Skene , Keaton J. Burns This is my paper

Pith reviewed 2026-05-19 13:25 UTC · model grok-4.3

classification 🧮 math.NA astro-ph.SRcs.NAphysics.ao-phphysics.flu-dyn
keywords automatic differentiationadjoint methodsspectral methodspartial differential equationsnumerical optimizationinverse problemsgradient computation
0
0 comments X

The pith

Reverse-mode automatic differentiation on symbolic PDE graphs produces efficient adjoint solvers for spectral methods.

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

The paper develops an automated technique that applies reverse-mode automatic differentiation directly to the symbolic graph representation of partial differential equations inside a spectral solver. This constructs adjoint solvers that preserve the speed and memory efficiency of sparse spectral methods for time-dependent and nonlinear systems. A sympathetic reader would care because it removes the need to derive or code adjoints by hand, making gradient-based optimization and sensitivity analysis practical across many equations, geometries, and boundary conditions.

Core claim

Representing the PDE symbolically allows reverse-mode differentiation to build the corresponding adjoint solver automatically, yielding correct gradients for arbitrary user-specified equations and nonlinearities while retaining the performance of spectral discretizations.

What carries the argument

Symbolic graph representation of the PDE, on which reverse-mode differentiation is applied to generate the adjoint solver.

If this is right

  • Gradient computation becomes available for optimization in a wide range of time-dependent nonlinear systems.
  • Users can perform inverse problems and sensitivity analyses without writing additional code for adjoints.
  • The approach extends to many different equations, geometries, and boundary conditions.
  • Parallel execution with MPI remains efficient for large simulations.

Where Pith is reading between the lines

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

  • Similar symbolic-graph techniques could reduce manual work in other high-level PDE modeling environments.
  • Easier access to gradients might speed up data assimilation and parameter tuning in applied physics models.
  • The method could be tested on problems with moving boundaries or coupled multiphysics systems to check generality.

Load-bearing premise

The symbolic graph must fully and accurately capture every detail of the PDE including nonlinear terms and boundary conditions.

What would settle it

A finite-difference approximation of the gradient for a known nonlinear PDE that disagrees with the gradient produced by the automated adjoint.

Figures

Figures reproduced from arXiv: 2506.14792 by Calum S. Skene, Keaton J. Burns.

Figure 1
Figure 1. Figure 1: Depiction of the adjoint-looping process to perform efficient PDE-based optimization using automatic discrete [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Neutral stability curve for plane Poiseuille flow [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: From the time series, we see that for all mag [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 3
Figure 3. Figure 3: Optimal kinematic dynamo solutions in the ball using nonlinear optimization via adjoint looping. Top [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Resolvent analysis for turbulent pipe flow at Re = [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Phase sensitivity analysis of the FitzHugh [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Taylor remainders demonstrating the convergence of finite difference estimates of the adjoint to [PITH_FULL_IMAGE:figures/full_fig_p014_6.png] view at source ↗
read the original abstract

We present a general and automated approach for computing model gradients for PDE solvers built on sparse spectral methods, and implement this capability in the widely used open-source Dedalus framework. We apply reverse-mode automatic differentiation to symbolic graph representations of PDEs, efficiently constructing adjoint solvers that retain the speed and memory efficiency of this important class of modern numerical methods. This approach enables users to compute gradients and perform optimization for a wide range of time-dependent and nonlinear systems without writing additional code. The framework supports a broad class of equations, geometries, and boundary conditions, and runs efficiently in parallel using MPI. We demonstrate the flexibility and capabilities of this system using canonical problems from the literature, showing both strong performance and practical utility for a wide variety of inverse problems. By integrating automatic adjoints into a flexible high-level solver, our approach enables researchers to perform gradient-based optimization and sensitivity analyses in spectral simulations with ease and efficiency.

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 presents a general automated approach for computing adjoints of PDEs discretized with sparse spectral methods. It applies reverse-mode automatic differentiation directly to the symbolic graph representation of the PDE inside the Dedalus framework, thereby generating adjoint solvers that the authors claim retain the original speed and memory characteristics. The work asserts that this enables gradient-based optimization and sensitivity analysis for arbitrary time-dependent nonlinear systems, geometries, and boundary conditions without users writing additional code, and demonstrates the capability on several canonical problems from the literature while running in parallel via MPI.

Significance. If the central claim is correct, the contribution is significant: it lowers the barrier to inverse problems and optimization in a class of high-performance spectral solvers that are already widely used in fluid dynamics, astrophysics, and related fields. Integration into an established open-source package with existing parallel infrastructure is a practical strength. The emphasis on retaining the efficiency of the underlying sparse spectral discretization distinguishes the work from generic AD wrappers that often incur substantial overhead.

major comments (2)
  1. [§3.2] §3.2 (Symbolic representation of the discrete system): The correctness of the generated adjoints for general boundary conditions and time-dependent problems rests on the assumption that every operation—including BC enforcement, spectral transforms, and time-stepping—is fully encoded inside the differentiable symbolic graph. The manuscript provides no explicit verification, pseudocode, or counter-example test showing that operations occurring outside the graph (e.g., in low-level Cython or MPI layers) are either absent or correctly differentiated. This is load-bearing for the claim that the method works for “arbitrary” BCs and time-dependent systems rather than only the demonstrated canonical cases.
  2. [§5.3] §5.3 (Performance and memory results): The reported wall-clock and memory figures are compared only against forward-mode AD and finite differences. No direct comparison is given to hand-derived adjoints for the same discretizations, which is the relevant baseline for assessing whether the automated approach truly “retains” the efficiency of the original sparse spectral solver. Without this comparison, the efficiency claim cannot be evaluated quantitatively.
minor comments (2)
  1. [Figure 4] Figure 4 caption: the legend labels “forward” and “adjoint” but does not specify whether these timings include the cost of building the symbolic graph or only the subsequent solve; this should be clarified.
  2. [Abstract] The abstract states that the method “runs efficiently in parallel using MPI,” yet no scaling plot or strong-scaling table is referenced in the abstract or introduction; adding a brief pointer would improve readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their careful review and constructive comments on the manuscript. We address each major comment point by point below, indicating where we agree and what revisions will be incorporated.

read point-by-point responses

  1. Referee: [§3.2] §3.2 (Symbolic representation of the discrete system): The correctness of the generated adjoints for general boundary conditions and time-dependent problems rests on the assumption that every operation—including BC enforcement, spectral transforms, and time-stepping—is fully encoded inside the differentiable symbolic graph. The manuscript provides no explicit verification, pseudocode, or counter-example test showing that operations occurring outside the graph (e.g., in low-level Cython or MPI layers) are either absent or correctly differentiated. This is load-bearing for the claim that the method works for “arbitrary” BCs and time-dependent systems rather than only the demonstrated canonical cases.

    Authors: We agree that explicit documentation of the graph scope would strengthen the presentation. In Dedalus the symbolic graph constructed by the Problem class encodes the full discrete residual, with boundary conditions incorporated through the tau method or lifting within the basis objects, spectral transforms applied as part of operator evaluation, and time-stepping driven by repeated residual evaluations. Low-level Cython kernels realize these symbolic operators directly and introduce no external operations that escape the graph; MPI parallelism acts only on the distributed array layout and does not modify the differentiation path. We will revise §3.2 to include pseudocode of graph construction and differentiation together with a verification test for a time-dependent problem using mixed boundary conditions. revision: yes

  2. Referee: [§5.3] §5.3 (Performance and memory results): The reported wall-clock and memory figures are compared only against forward-mode AD and finite differences. No direct comparison is given to hand-derived adjoints for the same discretizations, which is the relevant baseline for assessing whether the automated approach truly “retains” the efficiency of the original sparse spectral solver. Without this comparison, the efficiency claim cannot be evaluated quantitatively.

    Authors: The referee is correct that hand-derived adjoints constitute the most direct efficiency benchmark. For the broad class of arbitrary nonlinear time-dependent systems supported by the framework, however, constructing and maintaining hand-coded adjoints is impractical, which motivates the automated approach. Our results demonstrate that the generated adjoints exhibit the theoretically predicted scaling of reverse-mode differentiation on the same sparse graph: memory usage independent of the number of time steps (under standard checkpointing) and wall-clock cost comparable to a single forward solve. We will expand §5.3 with a discussion of the prohibitive cost of manual adjoint derivation for general problems and reference prior literature on adjoint overhead in spectral methods. revision: partial

Circularity Check

0 steps flagged

No circularity: standard reverse-mode AD applied to pre-existing Dedalus symbolic graph

full rationale

The paper's central contribution is the engineering implementation of reverse-mode automatic differentiation on the symbolic PDE graph already maintained inside Dedalus. This directly yields adjoint solvers by the standard properties of AD; no new mathematical derivation, fitted parameter, or uniqueness theorem is invoked that reduces to the inputs by construction. The approach is self-contained as a software extension of an existing symbolic representation, with no load-bearing self-citations or ansatzes required for the claim that gradients can be obtained without extra user code for supported systems.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The method rests on the assumption that Dedalus's existing symbolic representation of PDE operators is sufficiently complete and differentiable for arbitrary user equations; no new free parameters or invented physical entities are introduced.

axioms (2)
  • domain assumption Reverse-mode automatic differentiation applied to the symbolic graph of a PDE operator produces a correct and efficient adjoint solver.
    This is the central premise that allows the automation claim; it is invoked throughout the description of the approach.
  • domain assumption The sparse spectral discretization and time-stepping in Dedalus can be represented exactly enough in the symbolic graph for differentiation to be valid.
    Required for the method to apply to the full range of supported equations and boundary conditions.

pith-pipeline@v0.9.0 · 5695 in / 1482 out tokens · 19440 ms · 2026-05-19T13:25:04.580370+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

92 extracted references · 92 canonical work pages · 1 internal anchor

  1. [1]

    Unified Form Language: A domain-specific language for weak formulations of partial differential equa- tions

    M. S. Alnaes, A. Logg, K. B. Ølgaard, M. E. Rognes, and G. N. Wells. “Unified Form Language: A domain-specific language for weak formulations of partial differential equa- tions”. In: ACM Transactions on Mathematical Software 40 (2014)

    work page 2014

  2. [2]

    The photometric variability of massive stars due to gravity waves excited by core convection

    E. H. Anders et al. “The photometric variability of massive stars due to gravity waves excited by core convection”. In: Nature Astronomy 7.10 (2023), pp. 1228–1234

    work page 2023

  3. [3]

    Implicit- explicit Runge-Kutta methods for time-dependent partial differential equations

    U. M. Ascher, S. J. Ruuth, and R. J. Spiteri. “Implicit- explicit Runge-Kutta methods for time-dependent partial differential equations”. In: Applied Numerical Mathematics 25.2 (1997). Special Issue on Time Integration, pp. 151–167

    work page 1997

  4. [4]

    Periodicity in optimal hierar- chical checkpointing schemes for adjoint computations

    G. Aupy and J. Herrmann. “Periodicity in optimal hierar- chical checkpointing schemes for adjoint computations”. In: Optimization Methods and Software 32.3 (2017), pp. 594– 624

    work page 2017

  5. [5]

    Op- timal Multistage Algorithm for Adjoint Computation

    G. Aupy, J. Herrmann, P. Hovland, and Y. Robert. “Op- timal Multistage Algorithm for Adjoint Computation”. In: SIAM Journal on Scientific Computing 38.3 (2016), pp. C232–C255

    work page 2016

  6. [6]

    JAX- Fluids 2.0: Towards HPC for differentiable CFD of compress- ible two-phase flows

    D. A. Bezgin, A. B. Buhendwa, and N. A. Adams. “JAX- Fluids 2.0: Towards HPC for differentiable CFD of compress- ible two-phase flows”. In: Computer Physics Communica- tions 308 (Mar. 2025), p. 109433

    work page 2025

  7. [7]

    JAX- Fluids: A fully-differentiable high-order computational fluid dynamics solver for compressible two-phase flows

    D. A. Bezgin, A. B. Buhendwa, and N. A. Adams. “JAX- Fluids: A fully-differentiable high-order computational fluid dynamics solver for compressible two-phase flows”. In:Com- puter Physics Communications 282 (Jan. 2023), p. 108527

    work page 2023

  8. [8]

    Bouziani and D

    N. Bouziani and D. A. Ham. Physics-driven machine learn- ing models coupling PyTorch and Firedrake . 2023. arXiv: 2303.06871

    work page arXiv 2023

  9. [9]

    J. P. Boyd. Chebyshev and Fourier spectral methods. Courier Corporation, 2001

    work page 2001

  10. [10]

    Bradbury et al

    J. Bradbury et al. JAX: composable transformations of Python+NumPy programs. Version 0.3.13. 2018

    work page 2018

  11. [11]

    Corner Cases of the Generalized Tau Method

    K. J. Burns, D. Fortunato, K. Julien, and G. M. Vasil. “Corner Cases of the Generalized Tau Method”. In: preprint (2022)

    work page 2022

  12. [12]

    Dedalus: A flexible framework for numerical simu- lations with spectral methods

    K. J. Burns, G. M. Vasil, J. S. Oishi, D. Lecoanet, and B. P. Brown. “Dedalus: A flexible framework for numerical simu- lations with spectral methods”. In:Physical Review Research 2.2, 023068 (Apr. 2020), p. 023068

    work page 2020

  13. [13]

    The optimal kinematic dynamo driven by steady flows in a sphere

    L. Chen, W. Herreman, K. Li, P. W. Livermore, J. W. Luo, and A. Jackson. “The optimal kinematic dynamo driven by steady flows in a sphere”. In: Journal of Fluid Mechanics 839 (2018), pp. 1–32

    work page 2018

  14. [14]

    Variational algorithms for analysis and assimilation of meteorological observations: theoretical aspects

    F.-X. L. Dimet and O. Talagrand. “Variational algorithms for analysis and assimilation of meteorological observations: theoretical aspects”. In: Tellus A 38A.2 (1986), pp. 97–110

    work page 1986

  15. [15]

    checkpoint schedules: schedules for incremental checkpointing of adjoint simulations

    D. I. Dolci, J. R. Maddison, D. A. Ham, G. Pallez, and J. Herrmann. “checkpoint schedules: schedules for incremental checkpointing of adjoint simulations”. In: Journal of Open Source Software 9.95 (2024), p. 6148

    work page 2024

  16. [16]

    Dresdner et al

    G. Dresdner et al. Learning to correct spectral methods for simulating turbulent flows . 2022

    work page 2022

  17. [17]

    A gyroscopic polynomial basis in the sphere

    A. C. Ellison, K. Julien, and G. M. Vasil. “A gyroscopic polynomial basis in the sphere”. In: Journal of Computa- tional Physics 460 (2022), p. 111170

    work page 2022

  18. [18]

    G. B. Ermentrout and D. Terman. Mathematical Founda- tions of Neuroscience . Vol. 35. Interdisciplinary Applied Mathematics. New York, NY: Springer, 2010

    work page 2010

  19. [19]

    Automated Derivation of the Adjoint of High-Level Tran- sient Finite Element Programs

    P. E. Farrell, D. A. Ham, S. W. Funke, and M. E. Rognes. “Automated Derivation of the Adjoint of High-Level Tran- sient Finite Element Programs”. In: SIAM Journal on Sci- entific Computing 35.4 (2013), pp. C369–C393

    work page 2013

  20. [20]

    Impulses and Physiological States in Theo- retical Models of Nerve Membrane

    R. Fitzhugh. “Impulses and Physiological States in Theo- retical Models of Nerve Membrane.” In: Biophys J 1.6 (July 1961), pp. 445–466

    work page 1961

  21. [21]

    Sur les ´ equations diff´ erentielles lin´ eaires ` a co- efficients p´ eriodiques

    G. Floquet. “Sur les ´ equations diff´ erentielles lin´ eaires ` a co- efficients p´ eriodiques”. In:Annales scientifiques de l’ ´Ecole Normale Superieure 12 (1883), pp. 47–88

    work page

  22. [22]

    Giannakoglou, E

    K. Giannakoglou, E. Papoutsis-Kiachagias, and K. Gkaragk- ounis. adjointOptimisationFoam, an OpenFOAM-based opti- misation tool. the Parallel CFD & Optimization Unit, School of Mechanical Engineering, National Technical University of Athens. June 2019

    work page 2019

  23. [23]

    An introduction to the adjoint approach to design

    M. B. Giles and N. A. Pierce. “An introduction to the adjoint approach to design”. In: Flow, turbulence and combustion 65 (2000), pp. 393–415

    work page 2000

  24. [24]

    Optimal wave- form for fast synchronization of airfoil wakes

    V. Godavarthi, Y. Kawamura, and K. Taira. “Optimal wave- form for fast synchronization of airfoil wakes”. In: Journal of Fluid Mechanics 976 (2023), R1

    work page 2023

  25. [25]

    D. A. Ham et al. Firedrake User Manual. First edition. Im- perial College London et al. May 2023

    work page 2023

  26. [26]

    The Tapenade Automatic Dif- ferentiation tool: principles, model, and specification

    L. Hascoet and V. Pascual. “The Tapenade Automatic Dif- ferentiation tool: principles, model, and specification”. In: ACM Transactions on Mathematical Software 39.3 (2013)

    work page 2013

  27. [27]

    H-Revolve: A Framework for Adjoint Computation on Synchronous Hierarchical Plat- forms

    J. Herrmann and G. P. Aupy. “H-Revolve: A Framework for Adjoint Computation on Synchronous Hierarchical Plat- forms”. In: ACM Trans. Math. Softw. 46.2 (June 2020)

    work page 2020

  28. [28]

    On the part of the motion of the lunar perigee which is a function of the mean motions of the sun and moon

    G. W. Hill. “On the part of the motion of the lunar perigee which is a function of the mean motions of the sun and moon”. In: Acta Mathematica 8.1 (1886), pp. 1–36

    work page

  29. [29]

    A quantitative description of membrane current and its application to conduction and excitation in nerve

    A. L. Hodgkin and A. F. Huxley. “A quantitative description of membrane current and its application to conduction and excitation in nerve.” In: The Journal of physiology 117.4 (Aug. 1952), pp. 500–544

    work page 1952

  30. [30]

    Φ Flow (PhiFlow): Differentiable Simulations for PyTorch, TensorFlow and Jax

    P. Holl and N. Thuerey. “Φ Flow (PhiFlow): Differentiable Simulations for PyTorch, TensorFlow and Jax”. In:Interna- tional Conference on Machine Learning . PMLR. 2024

    work page 2024

  31. [31]

    Frequency synchronization in a random oscillator network

    T. Ichinomiya. “Frequency synchronization in a random oscillator network”. In: Phys. Rev. E 70 (2 Aug. 2004), p. 026116

    work page 2004

  32. [32]

    Isaac, N

    T. Isaac, N. Petra, G. Stadler, and O. Ghattas. “Scalable and efficient algorithms for the propagation of uncertainty from data through inference to prediction for large-scale prob- lems, with application to flow of the Antarctic ice sheet”. In: Journal of Computational Physics 296 (2015), pp. 348– 368. 15

    work page 2015

  33. [33]

    Scaling behaviour and control of nuclear wrinkling

    J. A. Jackson et al. “Scaling behaviour and control of nuclear wrinkling”. In: Nature physics 19.12 (2023), pp. 1927–1935

    work page 2023

  34. [34]

    Aerodynamic Shape Optimization

    A. Jameson. “Aerodynamic Shape Optimization”. In: Com- putational Aerodynamics . Cambridge Aerospace Series. Cambridge University Press, 2022, pp. 465–495

    work page 2022

  35. [35]

    Adjoint-based phase reduction analysis of incompressible periodic flows

    Y. Kawamura, V. Godavarthi, and K. Taira. “Adjoint-based phase reduction analysis of incompressible periodic flows”. In: Phys. Rev. Fluids 7 (10 Oct. 2022), p. 104401

    work page 2022

  36. [36]

    Nonlinear Nonmodal Stability Theory

    R. Kerswell. “Nonlinear Nonmodal Stability Theory”. In: Annual Review of Fluid Mechanics 50.Volume 50, 2018 (2018), pp. 319–345

    work page 2018

  37. [37]

    Machine learning–accelerated computational fluid dynamics

    D. Kochkov, J. A. Smith, A. Alieva, Q. Wang, M. P. Brenner, and S. Hoyer. “Machine learning–accelerated computational fluid dynamics”. In: Proceedings of the National Academy of Sciences 118.21 (2021)

    work page 2021

  38. [38]

    APEBench: A Benchmark for Autoregressive Neural Em- ulators of PDEs

    F. Koehler, S. Niedermayr, R. Westermann, and N. Thuerey. “APEBench: A Benchmark for Autoregressive Neural Em- ulators of PDEs”. In: Advances in Neural Information Pro- cessing Systems (NeurIPS) 38 (2024)

    work page 2024

  39. [39]

    Kuramoto

    Y. Kuramoto. Chemical Oscillations, Waves, and Turbu- lence. Berlin: Springer-Verlag, 1984

    work page 1984

  40. [40]

    SIMSOPT: A flexible framework for stellarator optimization

    M. Landreman, B. Medasani, F. Wechsung, A. Giuliani, R. Jorge, and C. Zhu. “SIMSOPT: A flexible framework for stellarator optimization”. In: Journal of Open Source Soft- ware 6.65 (2021), p. 3525

    work page 2021

  41. [41]

    Connection between non- linear energy optimization and instantons

    D. Lecoanet and R. R. Kerswell. “Connection between non- linear energy optimization and instantons”. In: Phys. Rev. E 97 (1 Jan. 2018), p. 012212

    work page 2018

  42. [42]

    Tensor calculus in spherical coordinates us- ing Jacobi polynomials. Part-II: implementation and exam- ples

    D. Lecoanet, G. M. Vasil, K. J. Burns, B. P. Brown, and J. S. Oishi. “Tensor calculus in spherical coordinates us- ing Jacobi polynomials. Part-II: implementation and exam- ples”. In: Journal of Computational Physics: X 3 (2019), p. 100012

    work page 2019

  43. [43]

    Phase- reduction for synchronization of oscillating flow by pertur- bation on surrounding structure

    I. A. Loe, H. Nakao, Y. Jimbo, and K. Kotani. “Phase- reduction for synchronization of oscillating flow by pertur- bation on surrounding structure”. In: Journal of Fluid Me- chanics 911 (2021), R2

    work page 2021

  44. [44]

    A. Logg, G. Wells, and K.-A. Mardal. Automated Solution of Differential Equations by the Finite Element Method: The FEniCS Book. Vol. 84. Apr. 2011

    work page 2011

  45. [45]

    Analysis methods for numerical weather pre- diction

    A. C. Lorenc. “Analysis methods for numerical weather pre- diction”. In: Quarterly Journal of the Royal Meteorological Society 112.474 (1986), pp. 1177–1194

    work page 1986

  46. [46]

    Adjoint Equations in Stability Analysis

    P. Luchini and A. Bottaro. “Adjoint Equations in Stability Analysis”. In: Annual Review of Fluid Mechanics 46.Volume 46, 2014 (2014), pp. 493–517

    work page 2014

  47. [47]

    A Comparison of Automatic Differentiation and Continu- ous Sensitivity Analysis for Derivatives of Differential Equa- tion Solutions

    Y. Ma, V. Dixit, M. J. Innes, X. Guo, and C. Rackauckas. “A Comparison of Automatic Differentiation and Continu- ous Sensitivity Analysis for Derivatives of Differential Equa- tion Solutions”. In: 2021 IEEE High Performance Extreme Computing Conference (HPEC). 2021, pp. 1–9

    work page 2021

  48. [48]

    Step-based checkpointing with high-level algorithmic differentiation

    J. R. Maddison. “Step-based checkpointing with high-level algorithmic differentiation”. In: Journal of Computational Science 82 (2024), p. 102405

    work page 2024

  49. [49]

    Au- tomated Calculation of Higher Order Partial Differential Equation Constrained Derivative Information

    J. R. Maddison, D. N. Goldberg, and B. D. Goddard. “Au- tomated Calculation of Higher Order Partial Differential Equation Constrained Derivative Information”. In: SIAM Journal on Scientific Computing 41.5 (2019), pp. C417– C445

    work page 2019

  50. [50]

    Adjoint Methods as Design Tools in Thermoa- coustics

    L. Magri. “Adjoint Methods as Design Tools in Thermoa- coustics”. In: Applied Mechanics Reviews 71.2 (Mar. 2019)

    work page 2019

  51. [51]

    A robust, discrete-gradient descent procedure for optimisa- tion with time-dependent PDE and norm constraints

    P. M. Mannix, C. S. Skene, D. Auroux, and F. Marcotte. “A robust, discrete-gradient descent procedure for optimisa- tion with time-dependent PDE and norm constraints”. In: The SMAI Journal of computational mathematics 10 (2024), pp. 1–28

    work page 2024

  52. [52]

    Further observations on the mean velocity distri- bution in fully developed pipe flow

    B. J. McKeon, J. LI, W. Jiang, J. F. Morrison, and A. J. Smits. “Further observations on the mean velocity distri- bution in fully developed pipe flow”. In: Journal of Fluid Mechanics 501 (2004), pp. 135–147

    work page 2004

  53. [53]

    A critical-layer framework for turbulent pipe flow

    B. J. McKeon and A. S. Sharma. “A critical-layer framework for turbulent pipe flow”. In: Journal of Fluid Mechanics 658 (2010), pp. 336–382

    work page 2010

  54. [54]

    Perturbation theory of nonlinear, non-self-adjoint eigenvalue problems: Simple eigenvalues

    G. A. Mensah, A. Orchini, and J. P. Moeck. “Perturbation theory of nonlinear, non-self-adjoint eigenvalue problems: Simple eigenvalues”. In: Journal of Sound and Vibration 473 (2020), p. 115200

    work page 2020

  55. [55]

    dolfin- adjoint 2018.1: automated adjoints for FEniCS and Fire- drake

    S. K. Mitusch, S. W. Funke, and J. S. Dokken. “dolfin- adjoint 2018.1: automated adjoints for FEniCS and Fire- drake”. In: Journal of Open Source Software 4.38 (2019), p. 1292

    work page 2018

  56. [56]

    An Active Pulse Transmission Line Simulating Nerve Axon

    J. Nagumo, S. Arimoto, and S. Yoshizawa. “An Active Pulse Transmission Line Simulating Nerve Axon”. In: Proceedings of the IRE 50.10 (1962), pp. 2061–2070

    work page 1962

  57. [57]

    Phase-based control of periodic flows

    A. G. Nair, K. Taira, B. W. Brunton, and S. L. Brunton. “Phase-based control of periodic flows”. In: Journal of Fluid Mechanics 927 (2021), A30

    work page 2021

  58. [58]

    Phase reduction approach to synchronisation of nonlinear oscillators

    H. Nakao. “Phase reduction approach to synchronisation of nonlinear oscillators”. In: Contemporary Physics 57.2 (2016), pp. 188–214

    work page 2016

  59. [59]

    Noise-Induced Syn- chronization and Clustering in Ensembles of Uncoupled Limit-Cycle Oscillators

    H. Nakao, K. Arai, and Y. Kawamura. “Noise-Induced Syn- chronization and Clustering in Ensembles of Uncoupled Limit-Cycle Oscillators”. In: Phys. Rev. Lett. 98 (18 May 2007), p. 184101

    work page 2007

  60. [60]

    Iterative methods for Navier-Stokes inverse problems

    L. O’Connor et al. “Iterative methods for Navier-Stokes inverse problems”. In: Phys. Rev. E 109 (4 Apr. 2024), p. 045108

    work page 2024

  61. [61]

    A fast and well-conditioned spectral method

    S. Olver and A. Townsend. “A fast and well-conditioned spectral method”. In: siam REVIEW 55.3 (2013), pp. 462– 489

    work page 2013

  62. [62]

    A sparse spectral method on triangles

    S. Olver, A. Townsend, and G. Vasil. “A sparse spectral method on triangles”. In: SIAM Journal on Scientific Com- puting 41.6 (2019), A3728–A3756

    work page 2019

  63. [63]

    Accurate solution of the Orr–Sommerfeld sta- bility equation

    S. A. Orszag. “Accurate solution of the Orr–Sommerfeld sta- bility equation”. In: Journal of Fluid Mechanics 50.4 (1971), pp. 689–703. 16

    work page 1971

  64. [64]

    Stanford University Unstructured (SU 2): An open-source integrated computational environment for multi-physics simulation and design

    F. Palacios et al. “Stanford University Unstructured (SU 2): An open-source integrated computational environment for multi-physics simulation and design”. In: 51st AIAA Aerospace Sciences Meeting including the New Horizons Fo- rum and Aerospace Exposition. 2013

    work page 2013

  65. [65]

    An adjoint method for neoclassical stellarator optimization

    E. J. Paul, I. G. Abel, M. Landreman, and W. Dorland. “An adjoint method for neoclassical stellarator optimization”. In: Journal of Plasma Physics 85.5 (2019), p. 795850501

    work page 2019

  66. [66]

    A review of the adjoint-state method for com- puting the gradient of a functional with geophysical appli- cations

    R.-E. Plessix. “A review of the adjoint-state method for com- puting the gradient of a functional with geophysical appli- cations”. In: Geophysical Journal International 167.2 (Nov. 2006), pp. 495–503

    work page 2006

  67. [67]

    Pringle, D

    G. Pringle, D. Jones, S. Goswami, S. Narayanan, and D. Goldberg. Providing the ARCHER community with ad- joint modelling tools for high-performance oceanographic and cryospheric computation. Tech. rep. Oct. 2016

    work page 2016

  68. [68]

    Energy requirement for a working dynamo

    M. Proctor. “Energy requirement for a working dynamo”. In: Geophysical & Astrophysical Fluid Dynamics 109.6 (2015), pp. 611–614

    work page 2015

  69. [69]

    Universal Differential Equations for Scientific Machine Learning

    C. Rackauckas et al. Universal Differential Equations for Scientific Machine Learning. 2021. arXiv: 2001.04385

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  70. [70]

    Adaptive numerical simulations with Trixi.jl: A case study of Julia for scientific computing

    H. Ranocha, M. Schlottke-Lakemper, A. R. Winters, E. Faulhaber, J. Chan, and G. J. Gassner. “Adaptive numerical simulations with Trixi.jl: A case study of Julia for scientific computing”. In: Proceedings of the JuliaCon Conferences1.1 (2022), p. 77

    work page 2022

  71. [71]

    An invitation to resolvent analysis

    L. V. Rolandi, J. H. M. Ribeiro, C.-A. Yeh, and K. Taira. “An invitation to resolvent analysis”. In: Theoretical and Computational Fluid Dynamics 38.5 (2024), pp. 603–639

    work page 2024

  72. [72]

    P. J. Schmid and D. S. Henningson. Stability and Transition in Shear Flows . Vol. 142. Applied Mathematical Sciences. New York, NY: Springer-Verlag, Jan. 2001

    work page 2001

  73. [73]

    Nonmodal Stability Theory

    P. J. Schmid. “Nonmodal Stability Theory”. In: Annual Re- view of Fluid Mechanics 39.Volume 39, 2007 (2007), pp. 129– 162

    work page 2007

  74. [74]

    C. S. Skene, F. Marcotte, and S. M. Tobias. On nonlinear transitions, minimal seeds and exact solutions for the geody- namo. 2024. arXiv: 2411.05499

    work page arXiv 2024

  75. [75]

    Phase-reduction analysis of peri- odic thermoacoustic oscillations in a Rijke tube

    C. S. Skene and K. Taira. “Phase-reduction analysis of peri- odic thermoacoustic oscillations in a Rijke tube”. In:Journal of Fluid Mechanics 933 (2022), A35

    work page 2022

  76. [76]

    MultiStage Approaches for Op- timal Offline Checkpointing

    P. Stumm and A. Walther. “MultiStage Approaches for Op- timal Offline Checkpointing”. In: SIAM Journal on Scien- tific Computing 31.3 (2009), pp. 1946–1967

    work page 2009

  77. [77]

    Phase-response analysis of synchro- nization for periodic flows

    K. Taira and H. Nakao. “Phase-response analysis of synchro- nization for periodic flows”. In: Journal of Fluid Mechanics 846 (2018), R2

    work page 2018

  78. [78]

    Modal Analysis of Fluid Flows: An Overview

    K. Taira et al. “Modal Analysis of Fluid Flows: An Overview”. In: AIAA Journal 55.12 (2017), pp. 4013–4041

    work page 2017

  79. [79]

    Modal Analysis of Fluid Flows: Applications and Outlook

    K. Taira et al. “Modal Analysis of Fluid Flows: Applications and Outlook”. In: AIAA Journal 58.3 (2020), pp. 998–1022

    work page 2020

  80. [80]

    The role of the Hessian matrix in fitting models to measurements

    W. C. Thacker. “The role of the Hessian matrix in fitting models to measurements”. In: Journal of Geophysical Re- search: Oceans 94.C5 (1989), pp. 6177–6196

    work page 1989

Showing first 80 references.