Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Diffeomorphic Neural Operator Learning

T0 review · 4 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read This paper claims that purely advective, reversible evolution can be learned exactly by representing time as composition on the diffeomorphism group, turning the semigroup into a group and hardwiring relabelling symmetry.

desk verdict Promising geometric operator-learning idea — learn a diffeomorphism and step by composition — but the visible text omits the experiments and analysis that would support the advertised claims. read the letter →

arxiv 2508.06690 v1 pith:MXSYWCHN submitted 2025-08-08 math.NA cs.NA

Seth Taylor , Alex Bihlo , Jean-Christophe Nave This is my paper
classification math.NAcs.NA MSC 65M9968T07
keywords operatorlearningdiffeomorphismgroupactionrelabellingsymmetrycharacteristicmappingturbulenceforecastingnon-diffusivetransportstructure-preservingneuralnetworks
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper argues that a class of evolution equations—those whose dynamics are purely advective and reversible—can be learned by a neural operator that never acts directly on field values. Instead, the network learns a 'lift' of the current field into the group of diffeomorphisms of the spatial domain, and time evolution is the group action of that diffeomorphism on the field. Because diffeomorphisms compose, the forward-in-time semigroup becomes a full group: two steps are one composed diffeomorphism, and every map has an inverse. The paper shows on linear advection and two-dimensional turbulence that this structure hardwires the relabelling symmetry of the dynamics, conserves mass, avoids numerical diffusion, and produces correct sub-grid statistical scaling. If this holds, it offers a general template for building infinite-dimensional geometric structure into operator learning.

What carries the argument

The central object is the group of diffeomorphisms $\mathrm{Diff}(\Omega)$ of the spatial domain, with composition as the group operation, paired with its action on the space of fields. The learned lift is a map $u \mapsto \phi_u$ from field space into $\mathrm{Diff}(\Omega)$; the predicted evolution is $u_t = u_0 \circ \phi_{u_0,t}^{-1}$ (for scalar fields), and multi-step forecasts are compositions of the learned diffeomorphisms. This group action is what converts the semigroup of forward evolution into a group, enforces the relabelling symmetry as a hard constraint, and enables the conservative, non-diffusive, resolution-robust behavior.

What would settle it

Compose a learned operator trained on 2D turbulence: for a holdout initial condition $u_0$, compare the network's predicted map at time $t_1+t_2$ against the composition of its maps at $t_2 \circ t_1$; any consistent mismatch, or any nonzero mass error, falsifies the exact group-action representation. Likewise, evaluating the same operator on forced-dissipative turbulence at a Reynolds number where enstrophy is not conserved should break the claimed non-diffusivity and the expected scaling relations.

Watch

Extended reading notes

Core claim

The central discovery is that for dynamics of pure transport—fields carried along a smooth invertible flow—the time-$t$ evolution operator can be represented exactly as a group action: a learned map takes the initial field to a diffeomorphism, and the evolved field is the pullback (or pushforward) of the initial field under that diffeomorphism. Time stepping then happens by composing diffeomorphisms, $\phi_{t+s} = \phi_s \circ \phi_t$, so the forward semigroup becomes a genuine group with inverses. The paper claims this hard constraint preserves the relabelling symmetry of the dynamics, conserves mass, and avoids numerical diffusion, and that these properties persist across resolutions. The

Load-bearing premise

The load-bearing premise is that the true dynamics are exactly a relabelling symmetry—fields are transported by a smooth invertible flow with no diffusion, mixing, forcing, or sources—so the evolution operator is exactly a diffeomorphism group action.

Editorial extensions

If this is right

  • Because the evolution is represented on the diffeomorphism group, forecasts at multiple time steps are obtained by composing learned maps, so the operator can be evaluated at arbitrary time horizons without re-entering field space.
  • The relabelling symmetry is hardwired into the operator, so mass is conserved and sharp discontinuities are not numerically diffused, as demonstrated on the linear advection test with a discontinuous initial condition.
  • The representation is resolution-robust: the same learned operator can be evaluated at different grid resolutions, and the paper shows it recovers anticipated statistical scaling at sub-grid scales in two-dimensional turbulence.
  • The semigroup-to-group transformation makes the forward map invertible, giving a principled route to backward time integration and data assimilation with the same operator.
  • The geometric perspective provides a template for embedding other infinite-dimensional symmetries, such as time translation for energy conservation, into operator learning.

Reading between the lines

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

  • A natural next test is vector-valued fields: the pushforward action of a diffeomorphism on velocity or vorticity introduces Jacobian factors, so the exact relabelling symmetry will only survive if the network also learns these geometric factors.
  • The composition law suggests the same operator can be evaluated at arbitrary time horizons and even inverted for backward integration; the paper does not demonstrate this, but it is a direct corollary of the group structure.
  • For real-world dissipative systems, a hybrid with a learned metamorphosis term—a direction the paper names—could retain the hard constraint for the advective part while absorbing sources and diffusion, an avenue the authors leave open.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes a geometric operator learning method for evolution equations. Instead of learning the field-to-field evolution directly, the method learns a lift into the space of diffeomorphisms of the spatial domain and advances fields by the group action (composition/pushforward). The authors claim this hard-codes a relabelling symmetry, conserves mass, avoids numerical diffusion, is robust to resolution changes, and captures sub-grid statistical scaling in turbulent flows. The visible text contains the abstract, introduction, references, and a short linear-advection experiment; the method section, resolution analysis, and turbulent-flow experiments promised in the abstract are not present. The core theoretical claim is that a class of evolution semigroups can be transformed into a diffeomorphism group structure, but the class is never characterized and no proof or error bound is supplied.

Significance. The idea of combining operator learning with diffeomorphic registration and characteristic-mapping techniques is timely and potentially valuable. If the hard symmetry constraint can be enforced exactly, the approach would be a useful addition to structure-preserving neural operators, and the connection to large-deformation diffeomorphic metric mapping is a genuine conceptual bridge. The paper also makes falsifiable numerical predictions (non-diffusivity, conservative behavior, sub-grid scaling). However, these contributions are currently presented only as claims: the central construction is not visible, the assumptions behind the group representation are not stated, and the advertised experiments are absent. As submitted, the manuscript does not support verification of its main assertions.

major comments (4)
  1. [Abstract and §1] The exactness condition of the method is never defined. The abstract restricts to 'a class of evolution operators' but the class is not characterized anywhere in the visible text. A solution operator can be exactly represented by composition with diffeomorphisms only if the dynamics is purely advective/reversible. For dynamics with diffusion, forcing, or entropy production, no exact group action exists. The advertised hard relabelling-symmetry constraint, non-diffusivity, and conservation all follow from exactness; if the learned lift is only approximate, these properties are approximate and can degrade. The manuscript needs a theorem or precise assumption characterizing admissible evolution operators, plus a residual/error bound for operators outside that class. The only visible experiment, linear advection, lies inside the reversible class and therefore does not test this load-bearing
  2. [Figures 2–3 and experimental section] The abstract promises numerical experiments on turbulent fluid dynamics demonstrating conservative properties, non-diffusivity, and sub-grid statistical scaling. No such experiments appear in the visible text. The only quantitative-looking material is the linear advection out-of-distribution test, but it lacks error metrics, error bars, hyperparameters, and a precise comparison against the direct field-evolution baseline. Without the turbulence experiments and a resolution study, the central numerical claims of the paper cannot be assessed. The authors should provide the complete experimental protocol, quantitative convergence data, and statistical/scaling measurements.
  3. [Abstract, semigroup-to-group claim] The statement that the approach 'transforms the semigroup structure of the evolution operator into a corresponding group structure' is asserted without proof or discussion of its validity. A genuine dissipative semigroup is not invertible and generally cannot be embedded in a group of diffeomorphisms. Imposing group structure may introduce spurious backward-time behavior or non-physical reversibility. The paper should either restrict attention to operators whose solution operators are group actions, or prove an embedding/approximation theorem and analyze the effect on backward modes and long-time predictions.
  4. [Resolution properties] The abstract and introduction claim that the paper studies resolution properties and that the approach is robust across resolutions, but no resolution analysis is present. A discretized group action does not automatically represent the continuous action at all resolutions; interpolation error, numerical diffusion, and aliasing can appear when composing discrete maps. Unless the manuscript supplies a quantitative resolution study or a mathematical invariance theorem, the resolution-robustness claim remains unsupported.
minor comments (5)
  1. [Figures 1–3] The figures have no captions, making it difficult to interpret the panels. Please add descriptive captions and include quantitative axes (e.g., error norms, mass conservation values).
  2. [Abstract and §1] Language issue: 'allowing time stepping be performed' should be 'allowing time stepping to be performed'.
  3. [Experimental description] The linear advection experiment lacks essential details: neural network architecture, training set size, number of epochs, optimizer, loss function, and the exact definition of the out-of-distribution test. These should be reported for reproducibility.
  4. [References] Reference [37] contains a typo: 'Infinite dimentional' should be 'Infinite dimensional'. Some references also lack complete publication details; please standardize.
  5. [Overall structure] The manuscript as presented is not self-contained: sections on the method, the diffeomorphic lift, and the numerical experiments are missing. Please ensure the submission includes all sections and that the text flows continuously from the introduction to the results.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the diffeomorphic-group construction enforces its structure-preserving properties by architecture, and the empirical claims are tested on held-out data; self-citations are to numerical methods and are not load-bearing.

full rationale

The paper's central claims either follow by construction from the diffeomorphic group-action parameterization (hard relabelling-symmetry constraint, conservation, non-diffusivity in the linear advection test) or are verified on data not used for fitting (out-of-distribution advection, and the turbulent statistical scalings claimed in the abstract). No fitted parameter is relabeled as a prediction: the learned lift is trained to match field evolution and then evaluated on new initial data. The abstract explicitly restricts to 'a class of evolution operators', which is a scope condition rather than a circular assumption. The self-citations (e.g., refs [34]-[36], [46], [66]) point to prior characteristic mapping and LDDMM-style numerical methods that the present work extends; they are algorithmic antecedents, not unverified uniqueness theorems or ansatze that secretly assume the target result. The skeptic's concern that irreversible or dissipative dynamics cannot be exactly represented by a diffeomorphism group action is a correctness/scope risk, not a circularity, and no visible text claims exact representability outside the stated class. Because no equation in the provided fragment reduces to its own input, and no prediction is forced by construction, the circularity score is 0.

Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

The central claim rests on the assumption that a diffeomorphic lift exists for the target dynamics, the faithfulness of the discretized group action, and the expressivity of the neural network. The network weights are fitted to data and are the main free parameters.

free parameters (1)
  • Neural network weights of the learned lift and field update = learned from training data
    All predictions depend on these fitted weights; no convergence or optimality guarantee is given in the visible text.
assumptions (3)
  • domain assumption The target evolution operator can be represented by a group action of diffeomorphisms on field space.
    Abstract restricts to 'a class of evolution operators'; this is the central modeling assumption.
  • ad hoc to paper The discretized group action faithfully represents the continuous action at all resolutions.
    Resolution invariance is claimed, but no proof is visible in the supplied text.
  • standard math The neural network can approximate the required diffeomorphisms to the needed accuracy.
    Relies on universal approximation results; not proven specifically for diffeomorphism-constrained maps.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Diffeomorphic Neural Operator Learning." pith.science (2026). https://pith.science/paper/MXSYWCHN

@misc{pith2026250806690,
  author       = {Pith},
  title        = {Pith review of: Diffeomorphic Neural Operator Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MXSYWCHN}},
  note         = {Machine review of arXiv:2508.06690}
}
read the original abstract

We present an operator learning approach for a class of evolution operators using a composition of a learned lift into the space of diffeomorphisms of the domain and the group action on the field space. In turn, this transforms the semigroup structure of the evolution operator into a corresponding group structure allowing time stepping be performed through composition on the space of diffeomorphisms rather than in the field space directly. This results in a number of structure-preserving properties related to preserving a relabelling symmetry of the dynamics as a hard constraint. We study the resolution properties of our approach, along with its connection to the techniques of diffeomorphic image registration. Numerical experiments on forecasting turbulent fluid dynamics are provided, demonstrating its conservative properties, non-diffusivity, and ability to capture anticipated statistical scaling relations at sub-grid scales. Our method provides an example of geometric operator learning and indicates a clear performance benefit from leveraging a priori known infinite-dimensional geometric structure.

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Neural Shape Operator Surrogates -- Expression Rate Bounds

    cs.LG 2026-04 unverdicted novelty 6.0 of 10

    Neural and spectral operators can approximate shape-to-solution maps for families of elliptic and parabolic PDEs and BIEs with provable uniform error bounds derived from parametric holomorphy on a reference domain.

Reference graph

Works this paper leans on

74 extracted references · 53 canonical work pages · cited by 1 Pith paper

  1. [1]

    Deep learning

    Yann LeCun, Yoshua Bengio, and Geoffrey Hinton. Deep learning. nature, 521(7553):436–444, 2015

    2015

  2. [2]

    Imagenet classification with deep convolutional neural networks

    Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton. Imagenet classification with deep convolutional neural networks. Advances in neural information processing systems, 25, 2012

    2012

  3. [3]

    Language models are few-shot learners

    Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Language models are few-shot learners. Advances in neural information processing systems, 33:1877–1901, 2020

    1901

  4. [4]

    The era5 global reanalysis

    Hans Hersbach, Bill Bell, Paul Berrisford, Shoji Hirahara, Andr´as Hor´anyi, Joaqu´ın Mu˜noz-Sabater, Julien Nicolas, Carole Peubey, Raluca Radu, Dinand Schepers, et al. The era5 global reanalysis. Quarterly Journal of the Royal Meteorological Society, 146(730):1999–2049, 2020

    work page 1999

  5. [5]

    Accurate medium-range global weather forecasting with 3d neural networks

    Kaifeng Bi, Lingxi Xie, Hengheng Zhang, Xin Chen, Xiaotao Gu, and Qi Tian. Accurate medium-range global weather forecasting with 3d neural networks. Nature, 619(7970):533–538, 2023

    work page 2023

  6. [6]

    Fourcastnet: Accelerating global high- resolution weather forecasting using adaptive fourier neural operators

    Thorsten Kurth, Shashank Subramanian, Peter Harrington, Jaideep Pathak, Morteza Mardani, David Hall, Andrea Miele, Karthik Kashinath, and Anima Anandkumar. Fourcastnet: Accelerating global high- resolution weather forecasting using adaptive fourier neural operators. In Proceedings of the Platform for Advanced Scientific Computing Conference, pages 1–11, 2023

    work page 2023

  7. [7]

    Learning skillful medium-range global weather forecasting

    Remi Lam, Alvaro Sanchez-Gonzalez, Matthew Willson, Peter Wirnsberger, Meire Fortunato, Ferran Alet, Suman Ravuri, Timo Ewalds, Zach Eaton-Rosen, Weihua Hu, et al. Learning skillful medium-range global weather forecasting. Science, 382(6677):1416–1421, 2023

    work page 2023

  8. [8]

    Model reduction and neural networks for parametric pdes

    Kaushik Bhattacharya, Bamdad Hosseini, Nikola B Kovachki, and Andrew M Stuart. Model reduction and neural networks for parametric pdes. The SMAI journal of computational mathematics, 7:121–157, 2021

    work page 2021

Show all 74 references
  1. [9]

    Fourier neural operator for parametric partial differential equations

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

    2010 arXiv

  2. [10]

    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

  3. [11]

    Convolutional neural operators

    Bogdan Raonic, Roberto Molinaro, Tobias Rohner, Siddhartha Mishra, and Emmanuel de Bezenac. Convolutional neural operators. In ICLR 2023 Workshop on Physics for Machine Learning, 2023. 17

    2023

  4. [12]

    Kernel methods are competitive for operator learning

    Pau Batlle, Matthieu Darcy, Bamdad Hosseini, and Houman Owhadi. Kernel methods are competitive for operator learning. Journal of Computational Physics, 496:112549, 2024

    2024

  5. [13]

    Physics-informed machine learning: case studies for weather and climate modelling.Philosophical Transactions of the Royal Society A, 379(2194):20200093, 2021

    Karthik Kashinath, M Mustafa, Adrian Albert, JL Wu, C Jiang, Soheil Esmaeilzadeh, Kamyar Azizzade- nesheli, R Wang, Ashesh Chattopadhyay, A Singh, et al. Physics-informed machine learning: case studies for weather and climate modelling.Philosophical Transactions of the Royal S...

    2021

  6. [14]

    Peter H Lauritzen, NK-R Kevlahan, Thomas Toniazzo, C Eldred, T Dubos, A Gassmann, VE Larson, C Jablonowski, O Guba, B Shipway, et al. Reconciling and improving formulations for thermodynamics and conservation principles in earth system models (esms).Journal of Advances in Mode...

    2022

  7. [15]

    Achieving conservation of energy in neural network emulators for climate modeling

    Tom Beucler, Stephan Rasp, Michael Pritchard, and Pierre Gentine. Achieving conservation of energy in neural network emulators for climate modeling. arXiv preprint arXiv:1906.06622, 2019

    1906 arXiv

  8. [16]

    Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations

    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

  9. [17]

    Physics-informed neural operator for learning partial differential equations

    Zongyi Li, Hongkai Zheng, Nikola Kovachki, David Jin, Haoxuan Chen, Burigede Liu, Kamyar Aziz- zadenesheli, and Anima Anandkumar. Physics-informed neural operator for learning partial differential equations. ACM/JMS Journal of Data Science, 1(3):1–27, 2024

    2024

  10. [18]

    Embedding hard physical constraints in neural network coarse-graining of 3d turbulence

    Arvind T Mohan, Nicholas Lubbers, Daniel Livescu, and Michael Chertkov. Embedding hard physical constraints in neural network coarse-graining of 3d turbulence. arXiv preprint arXiv:2002.00021, 2020

    2002 arXiv

  11. [19]

    Enforcing physical constraints in cnns through differentiable pde layer

    Karthik Kashinath, Philip Marcus, et al. Enforcing physical constraints in cnns through differentiable pde layer. In ICLR 2020 Workshop on Integration of Deep Neural Models and Differential Equations, 2020

    2020

  12. [20]

    Incorporating symmetry into deep dynamics models for improved generalization

    Rui Wang, Robin Walters, and Rose Yu. Incorporating symmetry into deep dynamics models for improved generalization. arXiv preprint arXiv:2002.03061, 2020

    2002 arXiv

  13. [21]

    Geometric deep learning: going beyond euclidean data

    Michael M Bronstein, Joan Bruna, Yann LeCun, Arthur Szlam, and Pierre Vandergheynst. Geometric deep learning: going beyond euclidean data. IEEE Signal Processing Magazine, 34(4):18–42, 2017

    2017

  14. [22]

    The curse of dimensionality in operator learning

    Samuel Lanthaler and Andrew M Stuart. The curse of dimensionality in operator learning. arXiv preprint arXiv:2306.15924, 2023

    2023 arXiv

  15. [23]

    Approximation of infinitely differentiable multivariate functions is intractable

    Erich Novak and Henryk Wo´zniakowski. Approximation of infinitely differentiable multivariate functions is intractable. Journal of Complexity, 25(4):398–404, 2009

    2009

  16. [24]

    Data complexity estimates for operator learning

    Nikola B Kovachki, Samuel Lanthaler, and Hrushikesh Mhaskar. Data complexity estimates for operator learning. arXiv preprint arXiv:2405.15992, 2024

    2024 arXiv

  17. [25]

    Sparse polynomial approximation of high- dimensional functions, volume 25

    Ben Adcock, Simone Brugiapaglia, and Clayton G Webster. Sparse polynomial approximation of high- dimensional functions, volume 25. SIAM, 2022

    2022

  18. [26]

    Approximation of high-dimensional parametric pdes

    Albert Cohen and Ronald DeV ore. Approximation of high-dimensional parametric pdes. Acta Numerica, 24:1–159, 2015

    2015

  19. [27]

    Exponential convergence of deep operator networks for elliptic partial differential equations

    Carlo Marcati and Christoph Schwab. Exponential convergence of deep operator networks for elliptic partial differential equations. SIAM Journal on Numerical Analysis, 61(3):1513–1545, 2023

    2023

  20. [28]

    Error estimates for deeponets: A deep learning framework in infinite dimensions

    Samuel Lanthaler, Siddhartha Mishra, and George E Karniadakis. Error estimates for deeponets: A deep learning framework in infinite dimensions. Transactions of Mathematics and Its Applications, 6(1):tnac001, 2022

    2022

  21. [29]

    Structure-preserving operator learning

    Nacime Bouziani and Nicolas Boull ´e. Structure-preserving operator learning. arXiv preprint arXiv:2410.01065, 2024

    2024 arXiv

  22. [30]

    Finite element exterior calculus

    Douglas N Arnold. Finite element exterior calculus. SIAM, 2018

    2018

  23. [31]

    Lagrangian flow networks for conservation laws

    F Arend Torres, Marcello Massimo Negri, Marco Inversi, Jonathan Aellen, and V olker Roth. Lagrangian flow networks for conservation laws. arXiv preprint arXiv:2305.16846, 2023

    2023 arXiv

  24. [32]

    Synthetic lagrangian turbulence by generative diffusion models

    Tianyi Li, Luca Biferale, Fabio Bonaccorso, Martino Andrea Scarpolini, and Michele Buzzicotti. Synthetic lagrangian turbulence by generative diffusion models. Nature Machine Intelligence, pages 1–11, 2024. 18

    2024

  25. [33]

    Recent advances on machine learning for computational fluid dynamics: A survey

    Haixin Wang, Yadi Cao, Zijie Huang, Yuxuan Liu, Peiyan Hu, Xiao Luo, Zezheng Song, Wanjia Zhao, Jilin Liu, Jinan Sun, et al. Recent advances on machine learning for computational fluid dynamics: A survey. arXiv preprint arXiv:2408.12171, 2024

    2024 arXiv

  26. [34]

    A characteristic mapping method for the two-dimensional incompressible euler equations

    Xi-Yuan Yin, Olivier Mercier, Badal Yadav, Kai Schneider, and Jean-Christophe Nave. A characteristic mapping method for the two-dimensional incompressible euler equations. Journal of Computational Physics, 424:109781, 2021

    2021

  27. [35]

    A characteristic mapping method for the three- dimensional incompressible euler equations

    Xi-Yuan Yin, Kai Schneider, and Jean-Christophe Nave. A characteristic mapping method for the three- dimensional incompressible euler equations. Journal of Computational Physics, 477:111876, 2023

    2023

  28. [36]

    A projection-based characteristic mapping method for tracer transport on the sphere

    Seth Taylor and Jean-Christophe Nave. A projection-based characteristic mapping method for tracer transport on the sphere. Journal of Computational Physics, 477:111905, 2023

    2023

  29. [37]

    Infinite dimentional lie groups with applications to mathematical physics

    Rudolf Schmid. Infinite dimentional lie groups with applications to mathematical physics. 2004

    2004

  30. [38]

    An introduction to infinite-dimensional differential geometry, volume 202

    Alexander Schmeding. An introduction to infinite-dimensional differential geometry, volume 202. Cam- bridge University Press, 2022

    2022

  31. [39]

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

    Nikola Kovachki, Zongyi Li, Burigede Liu, Kamyar Azizzadenesheli, Kaushik Bhattacharya, Andrew Stuart, and Anima Anandkumar. Neural operator: Learning maps between function spaces with applications to pdes. Journal of Machine Learning Research, 24(89):1–97, 2023

    2023

  32. [40]

    Operator learning: Algorithms and analysis

    Nikola B Kovachki, Samuel Lanthaler, and Andrew M Stuart. Operator learning: Algorithms and analysis. arXiv preprint arXiv:2402.15715, 2024

    2024 arXiv

  33. [41]

    Towards stability of autore- gressive neural operators

    Michael McCabe, Peter Harrington, Shashank Subramanian, and Jed Brown. Towards stability of autore- gressive neural operators. arXiv preprint arXiv:2306.10619, 2023

    2023 arXiv

  34. [42]

    Representation equivalent neural operators: a framework for alias-free operator learning

    Francesca Bartolucci, Emmanuel de Bezenac, Bogdan Raonic, Roberto Molinaro, Siddhartha Mishra, and Rima Alaifari. Representation equivalent neural operators: a framework for alias-free operator learning. Advances in Neural Information Processing Systems, 36, 2024

    2024

  35. [43]

    Spectral neural operators

    VS Fanaskov and Ivan V Oseledets. Spectral neural operators. In Doklady Mathematics, pages 1–7. Springer, 2024

    2024

  36. [44]

    Super-resolution neural operator

    Min Wei and Xuesong Zhang. Super-resolution neural operator. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 18247–18256, 2023

    2023

  37. [45]

    Discretization-invariance? on the discretization mis- match errors in neural operators

    Wenhan Gao, Ruichen Xu, Yuefan Deng, and Yi Liu. Discretization-invariance? on the discretization mis- match errors in neural operators. In The Thirteenth International Conference on Learning Representations, 2025

    2025

  38. [46]

    A characteristic mapping method for incompressible hydrodynamics on a rotating sphere

    Seth Taylor and Jean-Christophe Nave. A characteristic mapping method for incompressible hydrodynamics on a rotating sphere. arXiv preprint arXiv:2302.01205, 2023

    2023 arXiv

  39. [47]

    A fully lagrangian advection scheme

    John C Bowman, Mohammad Ali Yassaei, and Anup Basu. A fully lagrangian advection scheme. Journal of Scientific Computing, 64(1):151–177, 2015

    2015

  40. [48]

    Computational analysis of lddmm for brain mapping

    Can Ceritoglu, Xiaoying Tang, Margaret Chow, Darian Hadjiabadi, Damish Shah, Timothy Brown, Muhammad H Burhanullah, Huong Trinh, John T Hsu, Katarina A Ament, et al. Computational analysis of lddmm for brain mapping. Frontiers in neuroscience, 7:151, 2013

    2013

  41. [49]

    On the manifold structure of the space of brain images

    Samuel Gerber, Tolga Tasdizen, Sarang Joshi, and Ross Whitaker. On the manifold structure of the space of brain images. In Medical Image Computing and Computer-Assisted Intervention–MICCAI 2009: 12th International Conference, London, UK, September 20-24, 2009, Proceedings, Par...

    2009

  42. [50]

    Diffeomorphisms groups and pattern matching in image analysis

    Alain Trouv ´e. Diffeomorphisms groups and pattern matching in image analysis. International journal of computer vision, 28:213–221, 1998

    1998

  43. [51]

    Variational problems on flows of diffeomorphisms for image matching

    Paul Dupuis, Ulf Grenander, and Michael I Miller. Variational problems on flows of diffeomorphisms for image matching. Quarterly of applied mathematics, pages 587–600, 1998

    1998

  44. [52]

    Geodesic shooting for computational anatomy

    Michael I Miller, Alain Trouv ´e, and Laurent Younes. Geodesic shooting for computational anatomy. Journal of mathematical imaging and vision, 24:209–228, 2006

    2006

  45. [53]

    Diffeomorphic registration using geodesic shooting and gauss–newton optimisation

    John Ashburner and Karl J Friston. Diffeomorphic registration using geodesic shooting and gauss–newton optimisation. NeuroImage, 55(3):954–967, 2011. 19

    2011

  46. [54]

    Diffeomorphic 3d image registration via geodesic shooting using an efficient adjoint calculation

    Franc ¸ois-Xavier Vialard, Laurent Risser, Daniel Rueckert, and Colin J Cotter. Diffeomorphic 3d image registration via geodesic shooting using an efficient adjoint calculation. International Journal of Computer Vision, 97:229–241, 2012

    2012

  47. [55]

    Shapes and diffeomorphisms, volume 171

    Laurent Younes. Shapes and diffeomorphisms, volume 171. Springer, 2010

    2010

  48. [56]

    Overview of the geometries of shape spaces and diffeomorphism groups

    Martin Bauer, Martins Bruveris, and Peter W Michor. Overview of the geometries of shape spaces and diffeomorphism groups. Journal of Mathematical Imaging and Vision, 50:60–97, 2014

    2014

  49. [57]

    Computing large deformation metric mappings via geodesic flows of diffeomorphisms

    M Faisal Beg, Michael I Miller, Alain Trouv´e, and Laurent Younes. Computing large deformation metric mappings via geodesic flows of diffeomorphisms. International journal of computer vision, 61:139–157, 2005

    2005

  50. [58]

    On completeness of groups of diffeomorphisms

    Martins Bruveris and Franc ¸ois-Xavier Vialard. On completeness of groups of diffeomorphisms. Journal of the European Mathematical Society, 19(5):1507–1544, 2017

    2017

  51. [59]

    On universal approximation and error bounds for fourier neural operators

    Nikola Kovachki, Samuel Lanthaler, and Siddhartha Mishra. On universal approximation and error bounds for fourier neural operators. Journal of Machine Learning Research, 22(290):1–76, 2021

    2021

  52. [60]

    On the regularity of the composition of diffeomorphisms, volume 226

    Hasan Inci, Thomas Kappeler, and Peter Topalov. On the regularity of the composition of diffeomorphisms, volume 226. American mathematical society, 2013

    2013

  53. [61]

    Fredholm properties of riemannian exponential maps on diffeo- morphism groups

    Gerard Misiołek and Stephen C Preston. Fredholm properties of riemannian exponential maps on diffeo- morphism groups. Inventiones mathematicae, 179(1):191–227, 2010

    2010

  54. [62]

    On the volume elements on a manifold

    J¨urgen Moser. On the volume elements on a manifold. Transactions of the American Mathematical Society, 120(2):286–294, 1965

    1965

  55. [63]

    Wavelet neural operator: a neural operator for parametric partial differential equations

    Tapas Tripura and Souvik Chakraborty. Wavelet neural operator: a neural operator for parametric partial differential equations. arXiv preprint arXiv:2205.02191, 2022

    2022 arXiv

  56. [64]

    Neural operators with localized integral and differential kernels

    Miguel Liu-Schiaffini, Julius Berner, Boris Bonev, Thorsten Kurth, Kamyar Azizzadenesheli, and An- ima Anandkumar. Neural operators with localized integral and differential kernels. arXiv preprint arXiv:2402.16845, 2024

    2024 arXiv

  57. [65]

    A method for stochastic optimization

    Kingma Da. A method for stochastic optimization. arXiv preprint arXiv:1412.6980, 2014

    2014 arXiv

  58. [66]

    The characteristic mapping method for the linear advection of arbitrary sets

    Olivier Mercier, Xi-Yuan Yin, and Jean-Christophe Nave. The characteristic mapping method for the linear advection of arbitrary sets. SIAM Journal on Scientific Computing, 42(3):A1663–A1685, 2020

    2020

  59. [67]

    Two-dimensional turbulence.Reports on Progress in Physics, 43(5):547, 1980

    Robert H Kraichnan and David Montgomery. Two-dimensional turbulence.Reports on Progress in Physics, 43(5):547, 1980

    1980

  60. [68]

    Two-dimensional turbulence

    Guido Boffetta and Robert E Ecke. Two-dimensional turbulence. Annual review of fluid mechanics , 44:427–451, 2012

    2012

  61. [69]

    The measurement and analysis of shapes: An application of hydrody- namics and probability theory

    James Benn and Stephen Marsland. The measurement and analysis of shapes: An application of hydrody- namics and probability theory. Annals of Global Analysis and Geometry, 62(1):47–70, 2022

    2022

  62. [70]

    Mixture of kernels and iterated semidirect product of diffeomorphisms groups

    Martins Bruveris, Laurent Risser, and Franc ¸ois-Xavier Vialard. Mixture of kernels and iterated semidirect product of diffeomorphisms groups. Multiscale Modeling & Simulation, 10(4):1344–1368, 2012

    2012

  63. [71]

    The euler-poincar´e theory of metamorphosis

    Darryl Holm, Alain Trouv´e, and Laurent Younes. The euler-poincar´e theory of metamorphosis. Quarterly of Applied Mathematics, 67(4):661–685, 2009

    2009

  64. [72]

    A remark on stirling’s formula

    Herbert Robbins. A remark on stirling’s formula. The American mathematical monthly, 62(1):26–29, 1955

    1955

  65. [73]

    Infinite dimensional analysis: a hitchhiker’s guide

    Charalambos D Aliprantis and Kim C Border. Infinite dimensional analysis: a hitchhiker’s guide. Springer, 2006

    2006

  66. [74]

    Nonlocality and nonlinearity implies universality in operator learning, 2024

    Samuel Lanthaler, Zongyi Li, and Andrew M Stuart. Nonlocality and nonlinearity implies universality in operator learning, 2024. URL https://arxiv. org/abs/2304.13221. 20

    2024 arXiv

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.