pith. machine review for the scientific record. sign in

arxiv: 2604.16842 · v1 · submitted 2026-04-18 · 🧮 math.NA · cs.LG· cs.NA· math.AP

Recognition: unknown

Singularity Formation: Synergy in Theoretical, Numerical and Machine Learning Approaches

Yixuan Wang
Authors on Pith no claims yet

Pith reviewed 2026-05-10 07:08 UTC · model grok-4.3

classification 🧮 math.NA cs.LGcs.NAmath.AP
keywords singularity formationpartial differential equationsmodulation conditionsblowup profilesenergy estimatesKeller-Segel equationmachine learningKolmogorov-Arnold networks
0
0 comments X

The pith

Enforcing vanishing modulation conditions around approximate blowup profiles proves singularity formation in nonlinear PDEs such as the 3D Keller-Segel equation.

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

The paper develops an analytical framework that converts numerical observations of blowup into rigorous proofs by imposing vanishing modulation conditions on small perturbations around approximate blowup profiles and controlling them with singularly weighted energy estimates. This method is applied to equations with intricate asymptotics, including the nonlinear heat equation and the complex Ginzburg-Landau equation, where it yields clean proofs of finite-time singularity. The same approach resolves the previously open question of whether solutions to the three-dimensional Keller-Segel equation with logistic damping blow up in finite time. The work additionally refines numerical solvers and introduces machine-learning architectures to locate and characterize blowup solutions more precisely.

Core claim

We introduce a robust analytical framework to simplify and systematize pen-and-paper proofs for simpler singular PDEs based on enforcing vanishing modulation conditions for perturbations around approximate blowup profiles, complemented by singularly weighted energy estimates; the framework is shown to work on the nonlinear heat equation, the complex Ginzburg-Landau equation, and establishes singularity formation in the 3D Keller-Segel equation with logistic damping.

What carries the argument

vanishing modulation conditions imposed on perturbations around approximate blowup profiles, which reduce the problem to controllable energy estimates

If this is right

  • Singularity formation is rigorously established for the nonlinear heat equation and the complex Ginzburg-Landau equation.
  • Solutions of the 3D Keller-Segel equation with logistic damping develop singularities in finite time.
  • Numerical profiles can be turned into analytical proofs of blowup once modulation conditions are enforced and energy estimates are closed.
  • Machine-learning tools improve the detection and precise characterization of candidate blowup solutions.

Where Pith is reading between the lines

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

  • The same modulation-plus-energy strategy may become useful for other reaction-diffusion systems once reliable numerical profiles are available.
  • Hybrid numerical-analytical pipelines of this type could eventually supply guidance for more difficult open problems whose asymptotics are only partially understood.
  • The emphasis on interpretable learned nonlinearities in the accompanying machine-learning component suggests a route toward extracting new analytical ansatzes directly from data.

Load-bearing premise

Approximate blowup profiles obtained from numerical simulation are sufficiently accurate that the modulation conditions can be rigorously verified without hidden inconsistencies or missing higher-order terms.

What would settle it

A global smooth solution to the 3D Keller-Segel equation with logistic damping that remains bounded for all positive times, or a numerical computation showing persistent regularity past any candidate blowup time, would falsify the singularity claim.

Figures

Figures reproduced from arXiv: 2604.16842 by Yixuan Wang.

Figure 2.1
Figure 2.1. Figure 2.1: Comparison of the profile to the approximate steady state [PITH_FULL_IMAGE:figures/full_fig_p047_2_1.png] view at source ↗
Figure 2.2
Figure 2.2. Figure 2.2: Plot of the residue multiplied by the rescaled time [PITH_FULL_IMAGE:figures/full_fig_p048_2_2.png] view at source ↗
Figure 2.3
Figure 2.3. Figure 2.3: Fitting the law of the normalization constants. Left: [PITH_FULL_IMAGE:figures/full_fig_p049_2_3.png] view at source ↗
Figure 2.4
Figure 2.4. Figure 2.4: Fitting the law of the normalization constants for 2D. [PITH_FULL_IMAGE:figures/full_fig_p050_2_4.png] view at source ↗
Figure 3.1
Figure 3.1. Figure 3.1: Multi-Layer Perceptrons (MLPs) vs. Kolmogorov-Arnold Networks [PITH_FULL_IMAGE:figures/full_fig_p054_3_1.png] view at source ↗
Figure 3.2
Figure 3.2. Figure 3.2: Fitting the function 𝑓 (𝑥1, 𝑥2, 𝑥3, 𝑥4) = exp( 1 2 (sin(𝜋(𝑥 2 1 +𝑥 2 2 ))+sin(𝜋(𝑥 2 3 + 𝑥 2 4 )))). (a) 3-Layer KAN admits smooth representations. (b) The 2-Layer KAN learns highly oscillatory representations. (c) The 3-layer KAN achieves lower losses and has a smaller train-test gap than the 2-layer KAN. Grid extension A spline can be made arbitrarily accurate to a target function as the grid can be mad… view at source ↗
Figure 3.3
Figure 3.3. Figure 3.3: KANs are interpretable for simple symbolic tasks. [PITH_FULL_IMAGE:figures/full_fig_p064_3_3.png] view at source ↗
Figure 3.4
Figure 3.4. Figure 3.4: Knot dataset. Supervised mode (left): we rediscover DeepMind’s three [PITH_FULL_IMAGE:figures/full_fig_p065_3_4.png] view at source ↗
Figure 3.5
Figure 3.5. Figure 3.5: Compare KANs to MLPs on five toy examples. KANs can almost [PITH_FULL_IMAGE:figures/full_fig_p066_3_5.png] view at source ↗
Figure 3.6
Figure 3.6. Figure 3.6: Image fitting task (a PDE solution from PDEBench [342]). KAN [PITH_FULL_IMAGE:figures/full_fig_p067_3_6.png] view at source ↗
Figure 3.7
Figure 3.7. Figure 3.7: The PDE example. We plot L2 squared and H1 squared losses between [PITH_FULL_IMAGE:figures/full_fig_p068_3_7.png] view at source ↗
Figure 3.8
Figure 3.8. Figure 3.8: 1D wave dataset, where the target function has equal amplitudes of [PITH_FULL_IMAGE:figures/full_fig_p072_3_8.png] view at source ↗
Figure 3.9
Figure 3.9. Figure 3.9: 1D wave dataset, where the target function has increasing amplitudes of [PITH_FULL_IMAGE:figures/full_fig_p072_3_9.png] view at source ↗
Figure 3.10
Figure 3.10. Figure 3.10: The Gaussian random field dataset. Training losses of MLP and KANs, [PITH_FULL_IMAGE:figures/full_fig_p074_3_10.png] view at source ↗
Figure 3.11
Figure 3.11. Figure 3.11: The Gaussian random field dataset. Test losses of MLP and KANs, [PITH_FULL_IMAGE:figures/full_fig_p075_3_11.png] view at source ↗
Figure 3.12
Figure 3.12. Figure 3.12: Solving PDEs. 𝐿 2 and 𝐻 1 losses of MLP and KAN with different frequencies of the solution. 3.6.1 Related works Kolmogorov-Arnold theorem and neural networks. The connection between the Kolmogorov-Arnold theorem (KAT) and neural networks is not new in the literature [294, 321, 337, 195, 222, 201, 206, 111, 177, 297], but the pathological behavior of inner functions makes KAT appear unpromising in practi… view at source ↗
Figure 3.13
Figure 3.13. Figure 3.13: Left: Notations of activations that flow through the network. Right: an [PITH_FULL_IMAGE:figures/full_fig_p079_3_13.png] view at source ↗
Figure 4.1
Figure 4.1. Figure 4.1: Illustration of oversampling domains. On the right, we use an edge [PITH_FULL_IMAGE:figures/full_fig_p092_4_1.png] view at source ↗
Figure 4.2
Figure 4.2. Figure 4.2: Two level mesh: a fraction For a given equation, we compute the reference solution 𝑢ref using the classical FEM on the fine mesh with a sufficiently small ℎ, which we choose to be ℎ = 1/1024. By a posteriori estimates, we can check that the fine mesh indeed resolves the corresponding problems; thus, the associated fine mesh solutions could serve as accurate reference solutions for all of our numerical ex… view at source ↗
Figure 4.3
Figure 4.3. Figure 4.3: Numerical results for the periodic example. Left: [PITH_FULL_IMAGE:figures/full_fig_p097_4_3.png] view at source ↗
Figure 4.4
Figure 4.4. Figure 4.4: Left: the contour of log10 𝐴 for the high contrast example; right: the contour of 𝐴 for the rough media example. conditions, i.e., Γ2 = ∅, with a non-constant right-hand side 𝑓 (𝑥) = 𝑥 4 1 − 𝑥 3 2 + 1. In this example, we illustrate the convergence rate w.r.t the contrast 𝑀. We take different 𝑀 using the coarse mesh size 𝐻 = 2 −5 and 𝑚 = 1, 2, ..., 7. The numerical results are shown in [PITH_FULL_IMAGE:… view at source ↗
Figure 4.5
Figure 4.5. Figure 4.5: Numerical results for the high contrast example. Left: [PITH_FULL_IMAGE:figures/full_fig_p098_4_5.png] view at source ↗
Figure 4.6
Figure 4.6. Figure 4.6: Numerical results for the mixed boundary and rough field example. [PITH_FULL_IMAGE:figures/full_fig_p100_4_6.png] view at source ↗
read the original abstract

This thesis develops numerical and theoretical approaches for understanding and analyzing singularity formation in Partial Differential Equations (PDEs). The singularity formation in the Navier-Stokes Equation (NSE) is famously challenging as one of the seven Clay Prize problems. Unlike simpler equations such as the Nonlinear Heat (NLH) or Keller-Segel (KS) equations, where formal asymptotics near blowup are better understood, the intrinsic complexity of NSE makes quantitative analytical treatment difficult, if not impossible, without numerical guidance. Building on numerical insights, we introduce a robust analytical framework to simplify and systematize pen-and-paper proofs for simpler singular PDEs. We present a novel approach based on enforcing vanishing modulation conditions for perturbations around approximate blowup profiles, complemented by singularly weighted energy estimates. We demonstrate the efficacy of our method on PDEs with complicated asymptotics, such as NLH and the Complex Ginzburg-Landau (CGL) equation, and address the open problem of singularity formation in the 3D KS equation with logistic damping. We develop and refine numerical approaches that facilitate deeper insights into singularity formation. We demonstrate that machine learning methods significantly enhance our capability to identify and characterize potential blowup solutions with high precision. We improve on existing Physics-Informed Neural Network (PINN) and Neural Operator (NO) frameworks. Moreover, we present a novel machine learning paradigm, the Kolmogorov-Arnold Network (KAN) architecture, whose interpretability and excellent scaling properties are achieved through learnable nonlinearities.

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 / 1 minor

Summary. The manuscript develops theoretical, numerical, and machine learning approaches to singularity formation in PDEs. It introduces an analytical framework that enforces vanishing modulation conditions for perturbations around approximate blowup profiles, combined with singularly weighted energy estimates, to systematize proofs for equations with complex asymptotics. The framework is applied to the nonlinear heat equation (NLH), complex Ginzburg-Landau (CGL) equation, and the open problem of singularity formation in the 3D Keller-Segel equation with logistic damping. It also refines numerical methods and advances ML techniques, including improvements to physics-informed neural networks (PINNs) and neural operators, plus a novel Kolmogorov-Arnold Network (KAN) architecture emphasizing interpretability and scaling.

Significance. If the modulation-based framework converts numerical insights into independent rigorous proofs without hidden inconsistencies in the profiles or estimates, it would provide a valuable systematization for pen-and-paper analysis of singularity formation in singular PDEs and could resolve the open 3D KS problem. The ML components, especially KAN's learnable nonlinearities, would strengthen the synergy between simulation and theory in the field.

major comments (2)
  1. [Abstract/Introduction] Abstract and introduction: The central claim that the framework addresses the open problem of singularity formation in the 3D KS equation with logistic damping rests on enforcing vanishing modulation conditions and singularly weighted energy estimates, but the available text provides no derivations, error bounds, profile constructions, or verification steps, rendering the claim unverifiable and the numerical-to-rigorous conversion step unexamined for inconsistencies.
  2. [Framework section] Framework description: The assertion that numerical insights can be converted into rigorous analytical proofs via modulation conditions around approximate blowup profiles lacks explicit benchmarks or independent checks that the conditions hold without depending on fitted profiles, creating a moderate risk of circularity in the proofs for NLH, CGL, and KS.
minor comments (1)
  1. [Abstract] The abstract mentions 'complicated asymptotics' for NLH and CGL but does not specify which features of the asymptotics are addressed by the modulation conditions.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their thorough review and valuable comments on our work. We address each major comment below and outline the revisions we will make to strengthen the manuscript.

read point-by-point responses

  1. Referee: [Abstract/Introduction] Abstract and introduction: The central claim that the framework addresses the open problem of singularity formation in the 3D KS equation with logistic damping rests on enforcing vanishing modulation conditions and singularly weighted energy estimates, but the available text provides no derivations, error bounds, profile constructions, or verification steps, rendering the claim unverifiable and the numerical-to-rigorous conversion step unexamined for inconsistencies.

    Authors: We agree that the abstract and introduction provide a high-level summary without the full technical details. The derivations, error bounds, profile constructions, and verification steps for the 3D Keller-Segel equation are developed in detail in the body of the thesis, particularly in the sections on the analytical framework and its application to KS. To address this, we will revise the abstract and introduction to include a brief outline of the key steps in the proof, such as the construction of the approximate profile from numerical data and the subsequent verification of modulation conditions via energy estimates. This will make the conversion from numerical insights to rigorous proof more transparent and verifiable. revision: yes

  2. Referee: [Framework section] Framework description: The assertion that numerical insights can be converted into rigorous analytical proofs via modulation conditions around approximate blowup profiles lacks explicit benchmarks or independent checks that the conditions hold without depending on fitted profiles, creating a moderate risk of circularity in the proofs for NLH, CGL, and KS.

    Authors: The framework separates the numerical construction of approximate profiles from the analytical verification. Numerical methods, including the refined PINNs and KANs, are used to generate candidate profiles, but the modulation conditions are then enforced and verified through singularly weighted energy estimates that do not depend on the specific fitting procedure. To eliminate any perceived circularity, we will add explicit benchmarks in the revised manuscript, such as independent numerical checks of the modulation parameters and error estimates for NLH and CGL, and extend this to the KS case. This will demonstrate that the conditions hold rigorously. revision: yes

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The provided abstract and context describe an analytical framework that enforces vanishing modulation conditions around approximate blowup profiles together with singularly weighted energy estimates, then applies it to NLH, CGL, and the 3D KS equation. No specific equations, profile constructions, or derivation steps are quoted that reduce a claimed prediction or theorem to a fitted numerical input or self-citation by construction. The numerical insights are used as motivation for choosing profiles, but the modulation conditions and energy estimates are presented as independent analytical tools. Without a load-bearing reduction exhibited in the text, the derivation remains self-contained against external benchmarks and receives an honest non-finding.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

No specific free parameters, axioms, or invented entities are detailed in the abstract; the framework implicitly assumes existence of approximate blowup profiles from numerics and standard PDE regularity assumptions.

pith-pipeline@v0.9.0 · 5572 in / 1168 out tokens · 35065 ms · 2026-05-10T07:08:12.046040+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

300 extracted references · 99 canonical work pages · 8 internal anchors

  1. [1]

    The heterogeneous multiscale method

    Assyr Abdulle et al. “The heterogeneous multiscale method”. In:Acta Nu- merica21 (2012), pp. 1–87

    2012

  2. [2]

    Deepokan: Deepoperatornetworkbasedonkolmogorovarnoldnetworksformechanics problems

    Diab W Abueidda, Panos Pantidis, and Mostafa E Mobasher. “Deepokan: Deepoperatornetworkbasedonkolmogorovarnoldnetworksformechanics problems”. In:arXiv preprint arXiv:2405.19143(2024)

    work page arXiv 2024

  3. [3]

    Elsevier, 2003

    Robert A Adams and John JF Fournier.Sobolev spaces. Elsevier, 2003

    2003

  4. [4]

    Neural additive models: Interpretable machine learning with neural nets

    Rishabh Agarwal et al. “Neural additive models: Interpretable machine learning with neural nets”. In:Advances in neural information processing systems34 (2021), pp. 4699–4711

    2021

  5. [5]

    fKAN: Fractional Kolmogorov-Arnold networks with trainable Jacobi basis functions,

    Alireza Afzal Aghaei. “fKAN: Fractional Kolmogorov-Arnold Networks withtrainableJacobibasisfunctions”.In:arXivpreprintarXiv:2406.07456 (2024)

    work page arXiv 2024

  6. [6]

    rKAN:RationalKolmogorov-ArnoldNetworks

    AlirezaAfzalAghaei.“rKAN:RationalKolmogorov-ArnoldNetworks”.In: arXiv preprint arXiv:2406.14495(2024)

    work page arXiv 2024

  7. [7]

    GraphKAN: Graph Kol- mogorov Arnold Network for Small Molecule-Protein Interaction Predic- tions

    Tashin Ahmed and Md Habibur Rahman Sifat. “GraphKAN: Graph Kol- mogorov Arnold Network for Small Molecule-Protein Interaction Predic- tions”. In:ICML’24 Workshop ML for Life and Material Science: From Theory to Industry Applications. 2024

    2024

  8. [8]

    Residual-based attention in physics- informed neural networks

    Sokratis J Anagnostopoulos et al. “Residual-based attention in physics- informed neural networks”. In:Computer Methods in Applied Mechanics and Engineering421 (2024), p. 116805

    2024

  9. [9]

    URL: https://arxiv.org/abs/2406.09087, arXiv:2406.09087

    BasimAzamandNaveedAkhtar.“SuitabilityofKANsforComputerVision: A preliminary investigation”. In:arXiv preprint arXiv:2406.09087(2024)

    work page arXiv 2024

  10. [10]

    Deep spline networks with control ofLipschitzregularity

    Shayan Aziznejad and Michael Unser. “Deep spline networks with control ofLipschitzregularity”.In:ICASSP2019-2019IEEEInternationalConfer- ence on Acoustics, Speech and Signal Processing (ICASSP). IEEE. 2019, pp. 3242–3246

    2019

  11. [11]

    Numerical experience with a class of self-scaling quasi- Newton algorithms

    M Al-Baali. “Numerical experience with a class of self-scaling quasi- Newton algorithms”. In:Journal of optimization theory and applications 96 (1998), pp. 533–553

    1998

  12. [12]

    Broy- den’s quasi-Newton methods for a nonlinear system of equations and un- constrained optimization: a review and open problems

    Mehiddin Al-Baali, Emilio Spedicato, and Francesca Maggioni. “Broy- den’s quasi-Newton methods for a nonlinear system of equations and un- constrained optimization: a review and open problems”. In:Optimization Methods and Software29.5 (2014), pp. 937–954. 90

    2014

  13. [13]

    Optimal local approximation spaces for generalizedfiniteelementmethodswithapplicationtomultiscaleproblems

    Ivo Babuška and Robert Lipton. “Optimal local approximation spaces for generalizedfiniteelementmethodswithapplicationtomultiscaleproblems”. In:Multiscale Modeling & Simulation9.1 (2011), pp. 373–406

    2011

  14. [14]

    Can a finite element method perform ar- bitrarily badly?

    Ivo Babuška and John Osborn. “Can a finite element method perform ar- bitrarily badly?” In:Mathematics of Computation69.230 (2000), pp. 443– 462

    2000

  15. [15]

    Generalizedfiniteelementmethods:their performance and their relation to mixed methods

    IvoBabuškaandJohnEOsborn.“Generalizedfiniteelementmethods:their performance and their relation to mixed methods”. In:SIAM Journal on Numerical Analysis20.3 (1983), pp. 510–536

    1983

  16. [16]

    IsthepollutioneffectoftheFEMavoidable for the Helmholtz equation considering high wave numbers?

    IvoBabuškaandStefanSauter.“IsthepollutioneffectoftheFEMavoidable for the Helmholtz equation considering high wave numbers?” In:SIAM Journal on numerical analysis34.6 (1997), pp. 2392–2423

    1997

  17. [17]

    Multiscale-Spectral GFEM and optimal oversampling

    Ivo Babuška et al. “Multiscale-Spectral GFEM and optimal oversampling”. In:Computer Methods in Applied Mechanics and Engineering364 (2020), p. 112960

    2020

  18. [18]

    arXiv preprint arXiv:2102.06701 , year=

    Yasaman Bahri et al. “Explaining neural scaling laws”. In:arXiv preprint arXiv:2102.06701(2021)

    work page arXiv 2021

  19. [19]

    Diffusions hypercontractives

    Dominique Bakry and Michel Émery. “Diffusions hypercontractives”. In: Seminaire de probabilités XIX 1983/84. Springer, 1985, pp. 177–206

    1983

  20. [20]

    Universal approximation bounds for superpositions of a sigmoidal function

    Andrew R Barron. “Universal approximation bounds for superpositions of a sigmoidal function”. In:IEEE Transactions on Information theory39.3 (1993), pp. 930–945

    1993

  21. [21]

    Nearly-tight VC-dimension and pseudodimension boundsforpiecewiselinearneuralnetworks

    Peter L Bartlett et al. “Nearly-tight VC-dimension and pseudodimension boundsforpiecewiselinearneuralnetworks”.In:JournalofMachineLearn- ing Research20.63 (2019), pp. 1–17

    2019

  22. [22]

    Seeing implicit neural representations as fourier series

    Nuri Benbarka, Timon Höfer, Andreas Zell, et al. “Seeing implicit neural representations as fourier series”. In:Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. 2022, pp. 2041–2050

    2022

  23. [23]

    Formal asymptotics of bubbling in the harmonic map heat flow

    Jan Bouwe van den Berg, John R King, and Josephus Hulshof. “Formal asymptotics of bubbling in the harmonic map heat flow”. In:SIAM Journal on Applied Mathematics63.5 (2003), pp. 1682–1717

    2003

  24. [24]

    AmollifiedensembleKalmanfilter

    KayBergemannandSebastianReich.“AmollifiedensembleKalmanfilter”. In:Quarterly Journal of the Royal Meteorological Society136.651 (2010), pp. 1636–1643

    2010

  25. [25]

    AnensembleKalman-Bucyfilterfor continuous data assimilation

    KayBergemannandSebastianReich.“AnensembleKalman-Bucyfilterfor continuous data assimilation”. In:Meteorologische Zeitschrift21.3 (2012), p. 213. 91

    2012

  26. [26]

    A rescaling algorithm for the numeri- cal calculation of blowing-up solutions

    Marsha Berger and Robert V Kohn. “A rescaling algorithm for the numeri- cal calculation of blowing-up solutions”. In:Communications on pure and applied mathematics41.6 (1988), pp. 841–863

    1988

  27. [27]

    A Conceptual Introduction to Hamiltonian Monte Carlo

    Michael Betancourt. “A conceptual introduction to Hamiltonian Monte Carlo”. In:arXiv preprint arXiv:1701.02434(2017)

    work page Pith review arXiv 2017

  28. [28]

    ThegeometricfoundationsofHamiltonianMonte Carlo

    MichaelBetancourtetal.“ThegeometricfoundationsofHamiltonianMonte Carlo”. In:Bernoulli23.4A (2017), pp. 2257–2298

    2017

  29. [29]

    Discovering parametric activa- tion functions

    Garrett Bingham and Risto Miikkulainen. “Discovering parametric activa- tion functions”. In:Neural Networks148 (2022), pp. 48–65

    2022

  30. [30]

    Convolutional kolmogorov-arnold networks,

    Alexander Dylan Bodner et al. “Convolutional Kolmogorov-Arnold Net- works”. In:arXiv preprint arXiv:2406.13155(2024)

    work page arXiv 2024

  31. [31]

    Learning activation functions in deep (spline) neural networks

    Pakshal Bohra et al. “Learning activation functions in deep (spline) neural networks”. In:IEEE Open Journal of Signal Processing1 (2020), pp. 295– 309

    2020

  32. [32]

    Two-scale coupling for precon- ditioned Hamiltonian Monte Carlo in infinite dimensions

    Nawaf Bou-Rabee and Andreas Eberle. “Two-scale coupling for precon- ditioned Hamiltonian Monte Carlo in infinite dimensions”. In:Stochastics and Partial Differential Equations: Analysis and Computations9.1 (2021), pp. 207–242

    2021

  33. [33]

    Geometric integrators and the Hamiltonian Monte Carlo method

    Nawaf Bou-Rabee and Jesús María Sanz-Serna. “Geometric integrators and the Hamiltonian Monte Carlo method”. In:Acta Numerica27 (2018), pp. 113–206

    2018

  34. [34]

    RandomizedHamiltonian MonteCarlo

    NawafBou-RabeeandJesúsMaríaSanz-Serna.“RandomizedHamiltonian MonteCarlo”.In:TheAnnalsofAppliedProbability27.4(2017),pp.2159– 2194

    2017

  35. [35]

    Wav-kan: Wavelet kolmogorov-arnold networks,

    Zavareh Bozorgasl and Hao Chen. “Wav-kan: Wavelet kolmogorov-arnold networks”. In:arXiv preprint arXiv:2405.12832(2024)

    work page arXiv 2024

  36. [36]

    SusanneCBrenner,LRidgwayScott,andLRidgwayScott.Themathemat- ical theory of finite element methods. Vol. 3. Springer, 2008

    2008

  37. [37]

    Kagnns: Kolmogorov-arnold net- worksmeetgraphlearning

    Roman Bresson et al. “Kagnns: Kolmogorov-arnold networks meet graph learning”. In:arXiv preprint arXiv:2406.18380(2024)

    work page arXiv 2024

  38. [38]

    Universality in blow-up for nonlinear heat equations

    J. Bricmont and A. Kupiainen. “Universality in blow-up for nonlinear heat equations”. In:Nonlinearity7.2 (1994), p. 539

    1994

  39. [39]

    CRC press, 2011

    Steve Brooks et al.Handbook of Markov Chain Monte Carlo. CRC press, 2011

    2011

  40. [40]

    Formation of shocks for 2D isentropic compressible Euler

    Tristan Buckmaster, Steve Shkoller, and Vlad Vicol. “Formation of shocks for 2D isentropic compressible Euler”. In:Communications on Pure and Applied Mathematics75.9 (2022), pp. 2069–2120. 92

    2022

  41. [41]

    Randomized local model order re- duction

    Andreas Buhr and Kathrin Smetana. “Randomized local model order re- duction”. In:SIAM journal on scientific computing40.4 (2018), A2120– A2151

    2018

  42. [42]

    Partial regularity of suitable weak solutionsoftheNavier-Stokesequations

    L Caffarelli, R Kohn, and L Nirenberg. “Partial regularity of suitable weak solutionsoftheNavier-Stokesequations”.In:CommunicationsonPureand Applied Mathematics35.6 (1982), pp. 771–831

    1982

  43. [43]

    Physics-informed neural networks (PINNs) for fluid mechanics: A review

    Shengze Cai et al. “Physics-informed neural networks (PINNs) for fluid mechanics: A review”. In:Acta Mechanica Sinica37.12 (2021), pp. 1727– 1738

    2021

  44. [44]

    Multi-scaledeepneuralnetworksforsolving high dimensional pdes

    WeiCaiandZhi-QinJohnXu.“Multi-scaledeepneuralnetworksforsolving high dimensional pdes”. In:arXiv preprint arXiv:1910.11710(2019)

    work page arXiv 1910

  45. [45]

    Onexplicit𝐿 2-convergencerateesti- mateforunderdampedLangevindynamics

    YuCao,JianfengLu,andLihanWang.“Onexplicit𝐿 2-convergencerateesti- mateforunderdampedLangevindynamics”.In:arXivpreprintarXiv:1908.04746 (2019)

    work page arXiv 1908

  46. [46]

    Consensus-basedsampling

    JACarrilloetal.“Consensus-basedsampling”.In:StudiesinAppliedMath- ematics148.3 (2022), pp. 1069–1140

    2022

  47. [47]

    An analytical framework for consensus-based global optimization method

    José A Carrillo et al. “An analytical framework for consensus-based global optimization method”. In:Mathematical Models and Methods in Applied Sciences28.06 (2018), pp. 1037–1066

    2018

  48. [48]

    Particle,kinetic,andhydrodynamicmodelsofswarm- ing

    JoséACarrilloetal.“Particle,kinetic,andhydrodynamicmodelsofswarm- ing”. In:Mathematical modeling of collective behavior in socio-economic and life sciences. Springer, 2010, pp. 297–336

    2010

  49. [49]

    Optimal friction matrix for underdamped Langevin sampling

    Martin Chak et al. “Optimal friction matrix for underdamped Langevin sampling”. In:arXiv preprint arXiv:2112.06844(2021)

    work page arXiv 2021

  50. [50]

    On the regularity of the De Gregorio model for the 3D Euler equations

    Jiajie Chen. “On the regularity of the De Gregorio model for the 3D Euler equations”.In:J.Eur.Math.Soc.,toappear,arXivpreprintarXiv:2107.04777 (2021)

    work page arXiv 2021

  51. [51]

    On the Slightly Perturbed De Gregorio Model on𝑆1

    Jiajie Chen. “On the Slightly Perturbed De Gregorio Model on𝑆1”. In: ArchiveforRationalMechanicsandAnalysis241.3(2021),pp.1843–1869

    2021

  52. [52]

    Remarks on the smoothness of the𝐶1,𝛼 asymptotically self- similarsingularityinthe3DEulerand2DBoussinesqequations

    Jiajie Chen. “Remarks on the smoothness of the𝐶1,𝛼 asymptotically self- similarsingularityinthe3DEulerand2DBoussinesqequations”.In:arXiv preprint arXiv:2309.00150(2023)

    work page arXiv 2023

  53. [53]

    Singularityformationandglobalwell-posednessforthegener- alized Constantin–Lax–Majda equation with dissipation

    JiajieChen.“Singularityformationandglobalwell-posednessforthegener- alized Constantin–Lax–Majda equation with dissipation”. In:Nonlinearity 33.5 (2020), p. 2502

    2020

  54. [54]

    Finite time blowup of 2D Boussinesq and 3D Euler equations with𝐶1,𝛼 velocity and boundary

    Jiajie Chen and Thomas Y Hou. “Finite time blowup of 2D Boussinesq and 3D Euler equations with𝐶1,𝛼 velocity and boundary”. In:Communications in Mathematical Physics383.3 (2021), pp. 1559–1667. 93

    2021

  55. [55]

    Stable nearly self-similar blowup of the 2D Boussinesq and 3D Euler equations with smooth data I: Analysis.arXiv preprint: arXiv:2210.07191v3, 2022

    JiajieChenandThomasYHou.“Stablenearlyself-similarblowupofthe2D Boussinesq and 3D Euler equations with smooth data”. In:arXiv preprint arXiv:2210.07191(2022)

    work page arXiv 2022

  56. [56]

    Stablenearlyself-similarblowupofthe2D Boussinesqand3DEulerequationswithsmoothdataII:rigorousnumerics

    JiajieChenandThomasYHou.“Stablenearlyself-similarblowupofthe2D Boussinesqand3DEulerequationswithsmoothdataII:rigorousnumerics”. In:Multiscale Modeling & Simulation23.1 (2025), pp. 25–130

    2025

  57. [57]

    On the finite time blowup of theDeGregoriomodelforthe3DEulerequations

    Jiajie Chen, Thomas Y Hou, and De Huang. “On the finite time blowup of theDeGregoriomodelforthe3DEulerequations”.In:Communicationson pure and applied mathematics74.6 (2021), pp. 1282–1350

    2021

  58. [58]

    On the stability of blowup solutions to the complex Ginzburg-Landau equation in 𝑅𝑑

    Jiajie Chen, Thomas Y Hou, Van Tien Nguyen, and Yixuan Wang. “On the stability of blowup solutions to the complex Ginzburg-Landau equation in 𝑅𝑑”. In:Annals of PDE11.2 (2025), p. 29.doi:10.1007/s40818-025- 00223-1

    work page doi:10.1007/s40818-025- 2025

  59. [59]

    Randomized sampling for basis function construction in generalizedfiniteelementmethods

    Ke Chen et al. “Randomized sampling for basis function construction in generalizedfiniteelementmethods”.In:MultiscaleModeling&Simulation 18.2 (2020), pp. 1153–1177

    2020

  60. [60]

    Multiscale elliptic PDE upscaling and function approximation via subsampled data

    Yifan Chen and Thomas Y Hou. “Multiscale elliptic PDE upscaling and function approximation via subsampled data”. In:Multiscale Modeling & Simulation20.1 (2022), pp. 188–219

    2022

  61. [61]

    Exponential convergence for multiscale linear elliptic PDEs via adaptive edge basis functions

    Yifan Chen, Thomas Y Hou, and Yixuan Wang. “Exponential convergence for multiscale linear elliptic PDEs via adaptive edge basis functions”. In: Multiscale Modeling & Simulation19.2 (2021), pp. 980–1010.doi:10 . 1137/20M1352922

    2021

  62. [62]

    Exponentially convergent multiscale finite element method

    Yifan Chen, Thomas Y Hou, and Yixuan Wang. “Exponentially convergent multiscale finite element method”. In:Communications on Applied Mathe- maticsandComputation6.2(2024),pp.862–878.doi:10.1007/s42967- 023-00260-2

    work page doi:10.1007/s42967- 2024

  63. [63]

    Exponentially convergent multiscale methods for 2d high frequency heterogeneous Helmholtz equa- tions

    Yifan Chen, Thomas Y Hou, and Yixuan Wang. “Exponentially convergent multiscale methods for 2d high frequency heterogeneous Helmholtz equa- tions”. In:Multiscale Modeling & Simulation21.3 (2023), pp. 849–883. doi:10.1137/22M1507802

    work page doi:10.1137/22m1507802 2023

  64. [64]

    TENG: Time-evolving natural gradient for solving PDEs withdeepneuralnetstowardmachineprecision

    Zhuo Chen et al. “TENG: Time-evolving natural gradient for solving PDEs withdeepneuralnetstowardmachineprecision”.In:arXivpreprintarXiv:2404.10771 (2024)

    work page arXiv 2024

  65. [65]

    Demonstrating the Efficacy of Kolmogorov-Arnold Net- works in Vision Tasks

    Minjong Cheon. “Demonstrating the Efficacy of Kolmogorov-Arnold Net- works in Vision Tasks”. In:arXiv preprint arXiv:2406.14916(2024)

    work page arXiv 2024

  66. [66]

    Kolmogorov-arnold network for satellite image classification in remote sensing,

    Minjong Cheon. “Kolmogorov-Arnold Network for Satellite Image Classi- fication in Remote Sensing”. In:arXiv preprint arXiv:2406.00600(2024). 94

    work page arXiv 2024

  67. [67]

    Constraint energy minimizing generalized multiscale finite element method

    Eric T Chung, Yalchin Efendiev, and Wing Tat Leung. “Constraint energy minimizing generalized multiscale finite element method”. In:Computer Methods in Applied Mechanics and Engineering339 (2018), pp. 298–319

    2018

  68. [68]

    Calibrate, emulate, sample

    Emmet Cleary et al. “Calibrate, emulate, sample”. In:Journal of Computa- tional Physics424 (2021), p. 109716

    2021

  69. [69]

    NonradialtypeIIblowupfortheenergy-supercriticalsemilinear heat equation

    C.Collot.“NonradialtypeIIblowupfortheenergy-supercriticalsemilinear heat equation”. In:Anal. PDE10.1 (2017), pp. 127–252.issn: 2157-5045. doi:10.2140/apde.2017.10.127.url:https://doi.org/10.2140/ apde.2017.10.127

    work page doi:10.2140/apde.2017.10.127.url:https://doi.org/10.2140/ 2017

  70. [70]

    StronglyanisotropictypeIIblowupat anisolatedpoint

    C.Collot,F.Merle,andP.Raphael.“StronglyanisotropictypeIIblowupat anisolatedpoint”.In:JournaloftheAMS(2019).doi:https://doi.org/ 10.1090/jams/941.url:https://arxiv.org/abs/1709.04941

    work page doi:10.1090/jams/941.url:https://arxiv.org/abs/1709.04941 2019

  71. [71]

    Dynamics near the ground state for the energy critical nonlinear heat equation in large dimensions

    C. Collot, F. Merle, and P. Raphaël. “Dynamics near the ground state for the energy critical nonlinear heat equation in large dimensions”. In:Comm. Math. Phys.352.1 (2017), pp. 215–285.issn: 0010-3616.doi:10.1007/ s00220 - 016 - 2795 - 4.url:https : // doi .org / 10. 1007 /s00220 - 016-2795-4

    2017

  72. [72]

    Refined Description and Stability for Singular Solutions of the 2D Keller-Segel System

    C. Collot et al. “Refined Description and Stability for Singular Solutions of the 2D Keller-Segel System”. In:Communications on Pure and Applied Mathematics75.7 (2022), pp. 1419–1516

    2022

  73. [73]

    Stablesingularityformation fortheinviscidprimitiveequations

    CharlesCollot,SlimIbrahim,andQuyuanLin.“Stablesingularityformation fortheinviscidprimitiveequations”.In:Annalesdel’InstitutHenriPoincaré C(2023)

    2023

  74. [74]

    OntheEulerequationsofincompressiblefluids

    PConstantin.“OntheEulerequationsofincompressiblefluids”.In:Bulletin of the American Mathematical Society44.4 (2007), pp. 603–621

    2007

  75. [75]

    Note on loss of regularity for solutions of the 3D in- compressible Euler and related equations

    Peter Constantin. “Note on loss of regularity for solutions of the 3D in- compressible Euler and related equations”. In:Commun. Math. Phys.104 (1986), pp. 311–326

    1986

  76. [76]

    Asimpleone-dimensional modelforthethree-dimensionalvorticityequation

    PeterConstantin,PeterDLax,andAndrewMajda.“Asimpleone-dimensional modelforthethree-dimensionalvorticityequation”.In:Communicationson pure and applied mathematics38.6 (1985), pp. 715–724

    1985

  77. [77]

    Blow-up for the incompressible 3d-euler equations with uniformc1, 1 2 −ϵ ∩ l2 force.arXiv preprint arXiv:2309.08495, 2023

    Diego Córdoba and Luis Martínez-Zoroa. “Blow-up for the incompressible 3D-Euler equations with uniform𝐶 1, 1 2 −𝜖 ∩𝐿 2 force”. In:arXiv preprint arXiv:2309.08495(2023)

    work page arXiv 2023

  78. [78]

    Finite time singularities to the 3D incompressible Euler equations for solutions inC 1,α ∩C ∞(R3\{0})∩L 2.arXiv preprint arXiv:2308.12197, 2023

    DiegoCórdoba,LuisMartínez-Zoroa,andFanZheng.“Finitetimesingular- ities to the 3D incompressible Euler equations for solutions in𝐶∞(𝑅 3\0) ∩ 𝐶1,𝛼 ∩𝐿 2”. In:arXiv preprint arXiv:2308.12197(2023)

    work page arXiv 2023

  79. [79]

    MCMC methods for functions: modifying old algo- rithms to make them faster

    Simon L Cotter et al. “MCMC methods for functions: modifying old algo- rithms to make them faster”. In:Statistical Science28.3 (2013), pp. 424– 446. 95

    2013

  80. [80]

    Interpretable Machine Learning for Science with PySR and SymbolicRegression.jl

    MilesCranmer.“Interpretablemachinelearningfor sciencewithPySRand SymbolicRegression. jl”. In:arXiv preprint arXiv:2305.01582(2023)

    work page internal anchor Pith review arXiv 2023

Showing first 80 references.