Pith. sign in

REVIEW 49 references

Learning Hamiltonian Dynamics at Scale: A Differential-Geometric Approach

T0 review · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read A reduced-order Hamiltonian neural network (RO-HNN) learns a symplectic low-dimensional embedding and the dynamics on it, enabling stable long-term prediction for high-dimensional Hamiltonian systems up to 600 DoF.

arxiv 2509.24627 v2 pith:WA34ENRX submitted 2025-09-29 cs.LG

Katharina Friedl , Noémie Jaquier , Alyx Liao , Danica Kragic This is my paper
classification cs.LG
keywords hamiltoniandynamicsnetworkneuralhigh-dimensionalro-hnnconservationlaws
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

Many physical systems, like a swinging pendulum or a falling cloth, obey Hamiltonian mechanics: their motion is driven by a single energy function, and the flow preserves that energy. Neural networks that bake this structure in, called Hamiltonian neural networks (HNNs), make physically plausible predictions, but they become impractical when the system has many moving parts because the network must handle all coordinates at once.

RO-HNN avoids this by learning a low-dimensional 'latent' version of the system. A special autoencoder maps the high-dimensional state into a low-dimensional space while preserving the symplectic structure, the geometric property that enforces energy conservation. This is done not by adding a penalty term, but by constructing the encoder and decoder from constrained layers so that the symplectic property holds exactly. On this small latent space, a second network learns the reduced Hamiltonian (the energy function), and trajectories are integrated with a symplectic integrator that also respects the geometry.

The authors test RO-HNN on a 15-DoF pendulum, a 90-DoF particle vortex, and a 600-DoF damped cloth. In all cases the reduced model predicts trajectories that stay stable and match the true motion, in some cases over much longer horizons than were shown during training. The main advantage is scalability: the high-dimensional system is only used at the input and output of the autoencoder; all learning happens in a space of 3 to 10 coordinates. The trade-off is that the system must actually be reducible: if no low-dimensional symplectic submanifold captures the dynamics, the method will fail.

Extended reading notes

Core claim

RO-HNN 'provides physically-consistent, stable, and generalizable predictions of complex high-dimensional dynamics, thereby effectively extending the scope of Hamiltonian neural networks to high-dimensional physical systems.' (Abstract and Sec. 1.) If true, the contribution is a scalable reduction method that enforces Hamiltonian structure by construction rather than by soft penalties.

Load-bearing premise

The load-bearing premise is that a low-dimensional symplectic submanifold exists that captures the high-dimensional dynamics, i.e., the system is reducible. This is explicitly assumed in Sec. 3 ('We focus on systems ... for which the solutions γ(t) of the FOM (1) can be accurately approximated by a substantially lower dimensional surrogate model') and is not validated for arbitrary high-dimensional systems; the paper's synthetic pendulum is constructed to satisfy it by design (a latent 3-DoF pendulum mapped through a symplectomorphism built from a constrained AE of the same family as the model), making that benchmark favorable.

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.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The method introduces no new physical entities. Its central claim depends on the reducibility assumption and on standard smoothness/approximation assumptions. Main free parameters are the latent dimension, loss scaling, and the learned integrator coupling weight.

free parameters (4)
  • latent dimension d = 3 (pendulum), 6 and 10 (vortex/cloth)
    Chosen per system; controls trade-off between latent expressivity and HNN trainability.
  • loss scaling lambda = 1 (pendulum), 1e3 (vortex), 1e4 (cloth)
    Weights latent prediction loss against reconstruction; set by hand.
  • Strang integrator coupling weight w = learned via SoftPlus during training
    Controls binding of the augmented Hamiltonians in the nonseparable integrator; optimized as part of the network.
  • weight decay gamma = not reported
    Regularization coefficient in loss (12); no value given in the paper.
assumptions (4)
  • domain assumption The full-order system is Hamiltonian with a canonical symplectic form and a global chart; observations are canonical coordinates (q,p).
    Sec. 3 restricts to systems (M, J^T, H); Darboux makes this local but not global for all systems.
  • domain assumption The dynamics are reducible to a low-dimensional symplectic submanifold.
    Explicitly assumed in Sec. 3; not validated for general systems.
  • domain assumption A neural network can approximate the reduced Hamiltonian and the integrator remains accurate over training horizons.
    Standard ML approximation assumption; no error bounds are given.
  • domain assumption The constrained AE layers (10) with biorthogonal weights and invertible activations satisfy the projection properties exactly on the data manifold.
    True by construction for encoder-decoder pairs, but relies on exact Riemannian optimization convergence.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Learning Hamiltonian Dynamics at Scale: A Differential-Geometric Approach." pith.science (2026). https://pith.science/paper/WA34ENRX

@misc{pith2026250924627,
  author       = {Pith},
  title        = {Pith review of: Learning Hamiltonian Dynamics at Scale: A Differential-Geometric Approach},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WA34ENRX}},
  note         = {Machine review of arXiv:2509.24627}
}
read the original abstract

Embedding physical intuition into network architectures allows the learning of dynamics that enforce fundamental properties, such as energy conservation laws, thereby leading to physically-plausible predictions. Yet, scaling these models to high-dimensional dynamical systems remains a significant challenge. This paper introduces Reduced-order Hamiltonian Neural Network (RO-HNN), a novel physics-inspired neural network that combines the conservation laws of Hamiltonian mechanics with the scalability of model order reduction. RO-HNN is built on two core components: a novel geometrically-constrained symplectic autoencoder that learns a low-dimensional, structure-preserving symplectic submanifold, and a geometric Hamiltonian neural network that models the dynamics on the submanifold. Our experiments demonstrate that RO-HNN provides physically-consistent, stable, and generalizable predictions of complex high-dimensional dynamics, thereby effectively extending the scope of Hamiltonian neural networks to high-dimensional physical systems.

Figures

Figures reproduced from arXiv: 2509.24627 by the authors.

Figure 1
Figure 1. Flowchart of the forward dynamics of the geometric RO-HNN. The geometrically [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Relative error and reconstructed trajectories of the [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Ablation of the latent HNN architecture ( [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (12 more)
Figure 4
Figure 4. Figure 4: Predicted ( , , ) vs ground truth ( , , ) positions of the particle vortex. Times beyond 10s are out of the training data distribution [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Predicted positions of the damped cloth with a RO-HNN with [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Illustration of the latent system used to obtain the dynamics of a [PITH_FULL_IMAGE:figures/full_fig_p020_6.png]
Figure 7
Figure 7. Figure 7: Reconstructed trajectories of the RO-HNN ( [PITH_FULL_IMAGE:figures/full_fig_p024_7.png]
Figure 8
Figure 8. Figure 8: Left: Ablation of the latent HNN architecture on a doubled training set size |D| = 6000 compared to [PITH_FULL_IMAGE:figures/full_fig_p024_8.png]
Figure 9
Figure 9. Figure 9: Predicted ( , , ) vs ground truth ( , , ) positions of the particle vortex. The dynamics are learned with RO-HNN with d = 6 and d = 10. Times beyond 10s are out of the training data distribution. reconstruction error and conserves energy best during integration, showca…
Figure 10
Figure 10. Figure 10: Predicted ( , , ) vs ground truth ( , , ) reduced positions of the particle vortex in the latent space of the RO-HNN with d = 6 and d = 10. Times beyond 10s are out of the training data distribution [PITH_FULL_IMAGE:figures/full_fig_p026_10.png]
Figure 11
Figure 11. Figure 11: Predicted positions of the damped cloth with RO-HNNs with [PITH_FULL_IMAGE:figures/full_fig_p026_11.png]
Figure 12
Figure 12. Figure 12: Predicted cloth positions and momenta for [PITH_FULL_IMAGE:figures/full_fig_p027_12.png]
Figure 13
Figure 13. Figure 13: Top: Ground truth ( ) and predicted latent energies for 6-dimensional RO-HNNs with latent dissipation matrix parametrized with a SPD network ( ), a Cholesky network ( ), and ground truth values ( ), and 10-dimensional RO-HNNs with latent dissipation matrix parametrize…
Figure 14
Figure 14. Figure 14: Mean and standard deviation of the relative reconstruction ( [PITH_FULL_IMAGE:figures/full_fig_p028_14.png]
Figure 15
Figure 15. Figure 15: Meand and standard deviation of the latent prediction ( [PITH_FULL_IMAGE:figures/full_fig_p028_15.png]

Review history (2 revisions) →

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

49 extracted references · 3 canonical work pages

  1. [1]

    Ralph Abraham and Jerrold E. Marsden. Foundations of Mechanics. Addison-Wesley Publishing Company, Inc., second edition, 1987

    1987

  2. [2]

    Optimization Algorithms on Matrix Manifolds

    Pierre-Antoine Absil, Robert Mahony, and Rodolphe Sepulchre. Optimization Algorithms on Matrix Manifolds. Princeton University Press, 2007. URL https://press.princeton.edu/absil

    2007

  3. [3]

    Riemannian adaptive optimization methods

    Gary Becigneul and Octavian-Eugen Ganea. Riemannian adaptive optimization methods. In Intl. Conf. on Learning Representations ( ICLR ) , 2019. URL https://openreview.net/forum?id=r1eiqi09K7

    2019

  4. [4]

    Geometric optimization for structure-preserving model reduction of hamiltonian systems

    Thomas Bendokat and Ralf Zimmermann. Geometric optimization for structure-preserving model reduction of hamiltonian systems. IFAC-PapersOnLine, 55 0 (20): 0 457--462, 2022. URL https://www.sciencedirect.com/science/article/pii/S2405896322013386. 10th Vienna International Conference on Mathematical Modelling MATHMOD 2022

    2022

  5. [5]

    Which priors matter? B enchmarking models for learning latent dynamics

    Aleksandar Botev, Andrew Jaegle, Peter Wirnsberger, Daniel Hennes, and Irina Higgins. Which priors matter? B enchmarking models for learning latent dynamics. In Proceedings of the Neural Information Processing Systems Track on Datasets and Benchmarks, volume 1, 2021. URL https://datasets-benchmarks-proceedings.neurips.cc/paper_files/paper/2021/file/f033ab...

    2021

  6. [6]

    An introduction to optimization on smooth manifolds

    Nicolas Boumal. An introduction to optimization on smooth manifolds. Cambridge University Press, 2023. URL http://www.nicolasboumal.net/book

    2023

  7. [7]

    Brunton, Joshua L

    Steven L. Brunton, Joshua L. Proctor, and J. Nathan Kutz. Discovering governing equations from data by sparse identification of nonlinear dynamical systems. Proceedings of the National Academy of Sciences, 113 0 (15): 0 3932--3937, 2016. doi:10.1073/pnas.1517384113

    doi:10.1073/pnas.1517384113 2016

  8. [8]

    Symplectic model reduction of H amiltonian systems on nonlinear manifolds and approximation with weakly symplectic autoencoder

    Patrick Buchfink, Silke Glas, and Bernard Haasdonk. Symplectic model reduction of H amiltonian systems on nonlinear manifolds and approximation with weakly symplectic autoencoder. SIAM Journal of Scientific Computing , 45: 0 A289--A311, 2023. doi:10.1137/21m1466657

    work page doi:10.1137/21m1466657 2023

Show all 49 references
  1. [9]

    Model reduction on manifolds: A differential geometric framework

    Patrick Buchfink, Silke Glas, Bernard Haasdonk, and Benjamin Unger. Model reduction on manifolds: A differential geometric framework. Physica D: Nonlinear Phenomena, 468: 0 134299, 2024. doi:10.1016/j.physd.2024.134299

    2024

  2. [10]

    Nathan Kutz, and Steven L

    Kathleen Champion, Bethany Lusch, J. Nathan Kutz, and Steven L. Brunton. Data-driven discovery of coordinates and governing equations. Proceedings of the National Academy of Sciences, 116 0 (45): 0 22445--22451, 2019. doi:10.1073/pnas.1906995116

    2019 doi

  3. [11]

    Neural symplectic form: Learning H amiltonian equations on general coordinate systems

    Yuhan Chen, Takashi Matsubara, and Takaharu Yaguchi. Neural symplectic form: Learning H amiltonian equations on general coordinate systems. In Neural Information Processing Systems ( NeurIPS ) , 2021. URL https://proceedings.neurips.cc/paper_files/paper/2021/file/8b519f198dd26...

    2021

  4. [12]

    Symplectic recurrent neural networks

    Zhengdao Chen, Jianyu Zhang, Martin Arjovsky, and L \'e on Bottou. Symplectic recurrent neural networks. In Intl. Conf. on Learning Representations ( ICLR ) , 2020. URL https://openreview.net/forum?id=BkgYPREtPr

    2020

  5. [13]

    Cranmer, Sam Greydanus, Stephan Hoyer, Peter W

    Miles D. Cranmer, Sam Greydanus, Stephan Hoyer, Peter W. Battaglia, David N. Spergel, and Shirley Ho. L agrangian neural networks. In ICLR Deep Differential Equations Workshop, 2020. URL https://arxiv.org/abs/2003.04630

    2020 arXiv

  6. [14]

    Fernandes, and Waldyr M

    Pedro Duarte, Rui L. Fernandes, and Waldyr M. Oliva. Dynamics of the attractor in the lotka–volterra equations. Journal of Differential Equations, 149 0 (1): 0 143--189, 1998. doi:10.1006/jdeq.1998.3443

    1998

  7. [15]

    H amiltonian -based neural ODE networks on the SE (3) manifold for dynamics learning and control

    Thai Duong and Nikolay Atanasov. H amiltonian -based neural ODE networks on the SE (3) manifold for dynamics learning and control. In Robotics: Science and Systems ( R:SS ) , 2021. doi:10.15607/RSS.2021.XVII.086

    2021 doi

  8. [16]

    Geometries and interpolations for symmetric positive definite matrices, pp.\ 85--113

    Aasa Feragen and Andrea Fuster. Geometries and interpolations for symmetric positive definite matrices, pp.\ 85--113. Mathematics and Visualization. Springer, 2017. doi:10.1007/978-3-319-61358-1_5

    2017 doi

  9. [17]

    A R iemannian framework for learning reduced-order L agrangian dynamics

    Katharina Friedl, No \'e mie Jaquier, Jens Lundell, Tamim Asfour, and Danica Kragic. A R iemannian framework for learning reduced-order L agrangian dynamics. In Intl. Conf. on Learning Representations ( ICLR ) , 2025. URL https://openreview.net/forum?id=RoN6M3i7gJ

    2025

  10. [18]

    H amiltonian neural networks

    Samuel Greydanus, Misko Dzamba, and Jason Yosinski. H amiltonian neural networks. In Neural Information Processing Systems ( NeurIPS ) , volume 32, 2019. URL https://proceedings.neurips.cc/paper_files/paper/2019/file/26cd8ecadce0d4efd6cc8a8725cbd1f8-Paper.pdf

    2019

  11. [19]

    Hamilton

    William R. Hamilton. On a general method in dynamics. Philosophical Transactions of the Royal Society, pp.\ 247–308, 1834. doi:10.1098/rstl.1834.0017

  12. [20]

    Sympnets: Intrinsic structure-preserving symplectic networks for identifying H amiltonian systems

    Pengzhan Jin, Zhen Zhang, Aiqing Zhu, Yifa Tang, and George Em Karniadakis. Sympnets: Intrinsic structure-preserving symplectic networks for identifying H amiltonian systems. Neural Networks, 132: 0 166--179, 2020. doi:10.1016/j.neunet.2020.08.017

    2020 doi

  13. [21]

    Riemann tensor neural networks: Learning conservative systems with physics-constrained networks

    Anas Jnini, Lorenzo Breschi, and Flavio Vella. Riemann tensor neural networks: Learning conservative systems with physics-constrained networks. In Intl. Conf. on Machine Learning ( ICML ) , 2025. URL https://openreview.net/forum?id=cPMhMoJLAx&noteId=I8LbqaDsgw

    2025

  14. [22]

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

    George Em Karniadakis, Ioannis G. Kevrekidis, Lu Lu, Paris Perdikaris, Sifan Wang, and Liu Yang. Physics-informed machine learning. Nature Reviews Physics, 3 0 (6): 0 422--440, 2021. doi:10.1038/s42254-021-00314-5

    2021 doi

  15. [23]

    Geoopt: Riemannian optimization in PyTorch

    Max Kochurov, Rasul Karimov, and Serge Kozlukov. Geoopt: Riemannian optimization in PyTorch . arXiv:2005.02819, 2020. URL https://github.com/geoopt/geoopt

    2005 arXiv

  16. [24]

    John M. Lee. Introduction to smooth manifolds. Springer, 2013. doi:10.1007/978-1-4419-9982-5

    2013 doi

  17. [25]

    Simulating Hamiltonian Dynamics

    Benedict Leimkuhler and Sebastian Reich. Simulating Hamiltonian Dynamics. Cambridge University Press, 2005. doi:10.1017/CBO9780511614118

    2005 doi

  18. [26]

    Neural autoencoder-based structure-preserving model order reduction and control design for high-dimensional physical systems

    Marco Lepri, Davide Bacciu, and Cosimo Della Santina. Neural autoencoder-based structure-preserving model order reduction and control design for high-dimensional physical systems. IEEE Control Systems Letters , 8: 0 133--138, 2024. doi:10.1109/LCSYS.2023.3344286

    2024

  19. [27]

    Harnessing the power of neural operators with automatically encoded conservation laws

    Ning Liu, Yiming Fan, Xianyi Zeng, Milan Klöwer, Lu Zhang, and Yue Yu. Harnessing the power of neural operators with automatically encoded conservation laws. In Intl. Conf. on Machine Learning ( ICML ) , 2024. URL https://proceedings.mlr.press/v235/liu24p.html

    2024

  20. [28]

    Combining physics and deep learning to learn continuous-time dynamics models

    Michael Lutter and Jan Peters. Combining physics and deep learning to learn continuous-time dynamics models. Intl. Journal of Robotics Research, 42 0 (3): 0 83--107, 2023. doi:10.1177/02783649231169492

    2023 doi

  21. [29]

    Otto, Gregory R

    Samuel E. Otto, Gregory R. Macchio, and Clarence W. Rowley. Learning nonlinear projections for reduced-order modeling of dynamical systems using constrained autoencoders. Chaos: An Interdisciplinary Journal of Nonlinear Science, 33 0 (11), 2023. doi:10.1063/5.0169688

    2023 doi

  22. [30]

    Symplectic model reduction of H amiltonian systems

    Liqian Peng and Kamran Mohseni. Symplectic model reduction of H amiltonian systems. SIAM Journal on Scientific Computing, 38 0 (1): 0 A1--A27, 2016. doi:10.1137/140978922

    2016 doi

  23. [31]

    A R iemannian framework for tensor computing

    Xavier Pennec, Pierre Fillard, and Nicholas Ayache. A R iemannian framework for tensor computing. International Journal of Computer Vision, 66 0 (1): 0 41--66, 2006. doi:10.1007/s11263-005-3222-z

    2006 doi

  24. [32]

    Hamiltonian fluid mechanics

    Rick Salmon. Hamiltonian fluid mechanics. Annual Review of Fluid Mechanics, 20: 0 225--256, 1988. doi:10.1146/annurev.fl.20.010188.001301

    1988

  25. [33]

    Wilhelmus H. A. Schilders, Henk A. Van der Vorst, and Joost Rommes. Model Order Reduction: Theory, Research Aspects and Applications, volume 13. Springer Verlag, 2008. doi:10.1007/978-3-540-78841-6

    2008 doi

  26. [34]

    Quantisierung als eigenwertproblem

    Erwin Schrödinger. Quantisierung als eigenwertproblem. Annalen der Physik, 384 0 (4): 0 361--376, 1926. doi:10.1002/andp.19263840404

    1926 doi

  27. [35]

    Preserving lagrangian structure in data-driven reduced-order modeling of large-scale dynamical systems

    Harsh Sharma and Boris Kramer. Preserving lagrangian structure in data-driven reduced-order modeling of large-scale dynamical systems. Physica D: Nonlinear Phenomena, 462: 0 134128, 2024. doi:10.1016/j.physd.2024.134128

    2024

  28. [36]

    Symplectic model reduction of H amiltonian systems using data-driven quadratic manifolds

    Harsh Sharma, Hongliang Mu, Patrick Buchfink, Rudy Geelen, Silke Glas, and Boris Kramer. Symplectic model reduction of H amiltonian systems using data-driven quadratic manifolds. Computer Methods in Applied Mechanics and Engineering, 417: 0 116402, 2023. doi:10.1016/j.cma.2023.116402

    2023

  29. [37]

    Najera-Flores, Michael D

    Harsh Sharma, David A. Najera-Flores, Michael D. Todd, and Boris Kramer. Lagrangian operator inference enhanced with structure-preserving machine learning for nonintrusive model reduction of mechanical systems. Computer Methods in Applied Mechanics and Engineering, 423: 0 1168...

    2024

  30. [38]

    Dissipative H amiltonian neural networks: Learning dissipative and conservative dynamics separately

    Andrew Sosanya and Sam Greydanus. Dissipative H amiltonian neural networks: Learning dissipative and conservative dynamics separately. arXiv preprint arXiv:2201.10085, 2022. doi:10.48550/arxiv.2201.10085

    2022 doi arXiv

  31. [39]

    Explicit symplectic approximation of nonseparable H amiltonians: Algorithm and long time performance

    Molei Tao. Explicit symplectic approximation of nonseparable H amiltonians: Algorithm and long time performance. Phys. Rev. E, 94: 0 043303, 2016. doi:10.1103/PhysRevE.94.043303

    2016 doi

  32. [40]

    Mujoco: A physics engine for model-based control

    Emanuel Todorov, Tom Erez, and Yuval Tassa. Mujoco: A physics engine for model-based control. In IEEE/RSJ Intl. Conf. on Intelligent Robots and Systems ( IROS ) , pp.\ 5026--5033, 2012. doi:10.1109/IROS.2012.6386109

    2012

  33. [41]

    Understanding and mitigating gradient flow pathologies in physics-informed neural networks

    Sifan Wang, Yujun Teng, and Paris Perdikaris. Understanding and mitigating gradient flow pathologies in physics-informed neural networks. SIAM Journal on Scientific Computing , 43 0 (5): 0 A3055--A3081, 2021. doi:10.1137/20m1318043

    2021 doi

  34. [42]

    Nonseparable symplectic neural networks

    Shiying Xiong, Yunjin Tong, Xingzhe He, Shuqi Yang, Cheng Yang, and Bo Zhu. Nonseparable symplectic neural networks. In Intl. Conf. on Learning Representations ( ICLR ) , 2020. URL https://openreview.net/forum?id=B5VvQrI49Pa

    2020

  35. [43]

    Dissipative SymODEN : Encoding H amiltonian dynamics with dissipation and control into deep learning

    Yaofeng Desmond Zhong, Biswadip Dey, and Amit Chakraborty. Dissipative SymODEN : Encoding H amiltonian dynamics with dissipation and control into deep learning. In ICLR Workshop on Integration of Deep Neural Models and Differential Equations ( DeepDiffEq ) , 2020 a . URL https...

    2020

  36. [44]

    Symplectic ODE -net: Learning H amiltonian dynamics with control

    Yaofeng Desmond Zhong, Biswadip Dey, and Amit Chakraborty. Symplectic ODE -net: Learning H amiltonian dynamics with control. In Intl. Conf. on Learning Representations ( ICLR ) , 2020 b . URL https://openreview.net/forum?id=ryxmb1rKDS

    2020

  37. [45]

    Extending L agrangian and H amiltonian neural networks with differentiable contact models

    Yaofeng Desmond Zhong, Biswadip Dey, and Amit Chakraborty. Extending L agrangian and H amiltonian neural networks with differentiable contact models. In Neural Information Processing Systems ( NeurIPS ) , volume 34, pp.\ 21910--21922, 2021. URL https://proceedings.neurips.cc/p...

    2021

  38. [46]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

  39. [47]

    @esa (Ref

    \@ifxundefined[1] #1\@undefined \@firstoftwo \@secondoftwo \@ifnum[1] #1 \@firstoftwo \@secondoftwo \@ifx[1] #1 \@firstoftwo \@secondoftwo [2] @ #1 \@temptokena #2 #1 @ \@temptokena \@ifclassloaded agu2001 natbib The agu2001 class already includes natbib coding, so you should ...

  40. [48]

    \@lbibitem[] @bibitem@first@sw\@secondoftwo \@lbibitem[#1]#2 \@extra@b@citeb \@ifundefined br@#2\@extra@b@citeb \@namedef br@#2 \@nameuse br@#2\@extra@b@citeb \@ifundefined b@#2\@extra@b@citeb @num @parse #2 @tmp #1 NAT@b@open@#2 NAT@b@shut@#2 \@ifnum @merge>\@ne @bibitem@firs...

  41. [49]

    ^iziK W;x^.^ m mx xi _ ׵^Sq× = ޟik_K; ڞ I||]˚h緯u ,gx^ضڃ0ل 4s ! q^`K )Okْ

    @open @close @open @close and [1] URL: #1 \@ifundefined chapter * \@mkboth \@ifxundefined @sectionbib * \@mkboth * \@mkboth\@gobbletwo \@ifclassloaded amsart * \@ifclassloaded amsbook * \@ifxundefined @heading @heading NAT@ctr thebibliography [1] @ \@biblabel @NAT@ctr \@bibset...

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.