pith. sign in

arxiv: 2606.15053 · v2 · pith:6J5R4YS4new · submitted 2026-06-13 · 💻 cs.LG · cs.NA· math.NA

Physics-conforming Latent Twins

Matthias Chung , Yutong Bu , Deepanshu Verma This is my paper

Pith reviewed 2026-06-27 04:55 UTC · model grok-4.3

classification 💻 cs.LG cs.NAmath.NA
keywords surrogate modelslatent dynamicsphysics-informed learningconstraint preservationscientific machine learningsolution operatorsinvariant preservationdissipative systems
0
0 comments X

The pith

Latent surrogate solution operators can be trained to obey selected physical conservation and dissipation rules by design through transferred constraints.

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

The paper introduces Physics-conforming Latent Twins as a way to build fast surrogate models for time-dependent physical systems that respect conservation laws, invariants, and dissipative structures instead of only matching training data. Standard latent models risk producing unphysical long-term behavior even when they interpolate trajectories accurately. The method jointly optimizes an encoder, decoder, and latent flow map while adding constraints in the latent space that correspond to physical structure in the original variables. It supplies a constraint-transfer viewpoint plus mathematical bounds that link latent enforcement to reduced physical defects after decoding, along with algebraic conditions the flow map must satisfy to preserve linear or quadratic invariants or enforce dissipation. Experiments on ODE and PDE benchmarks confirm better structural fidelity without loss of predictive accuracy.

Core claim

Physics-conforming Latent Twins jointly learn an encoder, a decoder, and a latent flow map between time-indexed states while constraining the latent dynamics to preserve or dissipate prescribed structural quantities; a constraint-transfer viewpoint connects physical structure in the original state space to compatible latent constraints, structure-preservation bounds quantify the resulting improvement in decoded fidelity, and algebraic conditions on the flow map guarantee preservation of linear and quadratic invariants or enforcement of dissipative inequalities.

What carries the argument

Constraint-transfer viewpoint that links physical structure in state space to compatible constraints in latent space, enabling enforcement inside the latent flow map.

If this is right

  • Latent enforcement of the transferred constraints produces lower physical defects in the decoded trajectories than unconstrained latent models.
  • Latent flow maps satisfying the derived algebraic conditions preserve linear and quadratic invariants or enforce the prescribed dissipative inequalities by construction.
  • Surrogate predictions maintain pointwise accuracy while exhibiting improved qualitative long-time behavior on representative ODE and PDE systems.
  • The same framework applies to both linear and nonlinear invariants provided the algebraic conditions for the flow map can be satisfied.

Where Pith is reading between the lines

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

  • The approach could be combined with existing reduced-order modeling techniques that already operate in latent spaces to add physics compliance at low extra cost.
  • Extension to systems with time-dependent or state-dependent constraints would require generalizing the algebraic conditions derived for fixed invariants.
  • If the latent dimension is chosen too small, the transferred constraints may become incompatible and force a trade-off between representation power and structure preservation.
  • The structure-preservation bounds suggest a quantitative way to choose the strength of the latent penalty term during training.

Load-bearing premise

Compatible structural constraints can be found and transferred into latent space without destroying the encoder-decoder pair's ability to represent the true dynamics accurately.

What would settle it

If numerical experiments on the ODE and PDE benchmarks show that enforcing the latent constraints produces no measurable reduction in physical defects after decoding, the structure-preservation bounds and the overall approach would be falsified.

Figures

Figures reproduced from arXiv: 2606.15053 by Deepanshu Verma, Matthias Chung, Yutong Bu.

Figure 1
Figure 1. Figure 1: Schematic of a Physics-conforming Latent Twin. The encoder 𝑒 maps a state 𝑢(𝑠) ∈  to a latent state 𝑧𝑠 ∈ , the latent flow 𝑚𝑠→𝑡 propagates 𝑧𝑠 to 𝑧𝑡 , and the decoder 𝑑 maps the latent states back to reconstructed or predicted states. The structural functional  encodes a physical property in the state space, for example a conserved quantity or a dissipative functional. A compatible latent constraint 𝑍 r… view at source ↗
Figure 2
Figure 2. Figure 2: Comparison of the Physics-conforming Latent Twin, POD, and DMD on the stable linear ODE 𝑥 ′ = 𝑀𝑥 with ambient dimension 𝑛𝑥 = 100 and intrinsic dimension 𝑟 = 10. POD provides an optimal projection benchmark in this constructed low-rank linear setting. Since the linear reconstruction map already yields a quadratic decoder-induced pullback, we set 𝜆pullback = 0 and enforce the compatible latent invariant dire… view at source ↗
Figure 3
Figure 3. Figure 3: 95% prediction-error bands for the SIR experiments. (a) Linear autoencoder: 𝑥𝑡 = 𝑃 ⊤ exp((𝑡 − 𝑠)𝑊 )𝑃 𝑥𝑠 . (b) Nonlinear autoencoder: 𝑥𝑡 = (𝑑◦ exp((𝑡 − 𝑠)𝑊 )◦𝑒)(𝑥𝑠 ). The nonlinear model reduces the error magnitude by approximately three orders of magnitude relative to the linear model. in Equations (7) and (8). We use (𝑥) = 𝟏 ⊤𝑥, 𝑍(𝑧) = 𝟏 ⊤ 𝑧, and set 𝜆pullback = 𝜆latent = 1. The pullback loss aligns the… view at source ↗
Figure 4
Figure 4. Figure 4: Relative prediction errors for the Physics-conforming Latent Twin, POD, and DMD on the multigroup SIR system with 𝑛𝑥 = 60 and reduced dimension 𝑛𝑧 = 3. The Latent Twin achieves lower error by combining a nonlinear state representation with pullback and latent conformity regularization that encode the population conservation structure. 𝑊 ⊤𝑄𝑍 + 𝑄𝑍𝑊 = 0. We impose this condition as a hard architectural constr… view at source ↗
Figure 5
Figure 5. Figure 5: Undamped pendulum predictions on an extended time interval. All models are trained using information from 𝑡 ∈ [0, 10] and evaluated on 𝑡 ∈ [0, 15]. The left column shows the predicted state trajectory and the right column shows the prediction error. The unconstrained Latent Twin loses accuracy under extrapolation because the learned latent generator is not structurally restricted. The energy-regularized PI… view at source ↗
Figure 6
Figure 6. Figure 6: Ablation study for the undamped pendulum. Mean absolute prediction errors and Hamiltonian defects are reported on the observed interval [0, 10] and the extrapolation interval [10, 15]. Pullback matching aligns the latent invariant with the decoder-induced Hamiltonian, while latent conformity enforces preservation by the latent flow. Their combination gives the most reliable extrapolation performance and sm… view at source ↗
Figure 7
Figure 7. Figure 7: Gradient-flow Rosenbrock example. Left: relative prediction error for the unconstrained Latent Twin Ψ( ̃ 𝑡, 𝑠, 𝑥𝑠 ) and the dissipative Latent Twin Ψ(𝑡, 𝑠, 𝑥𝑠 ), with the shaded region indicating the training horizon. Right: representative trajectory in the Rosenbrock landscape. Both models are accurate on the training range, while the dissipative model remains closer to the true trajectory under extrapola… view at source ↗
Figure 8
Figure 8. Figure 8: Heat equation predictions. Predicted temperature fields (row 1), true temperature fields (row 2), and absolute errors (row 3) at eight time snapshots over the horizon 𝑡 ∈ [0, 5]. The Physics-conforming Latent Twins accurately tracks the diffusive evolution, with small localized errors throughout the horizon. 4.3.1. Heat equation We next consider the two-dimensional heat equation 𝜕𝑡 𝑢 = 𝜅Δ𝑢, which describes… view at source ↗
Figure 9
Figure 9. Figure 9: Conservation diagnostics for the heat Latent Twin. Top: 𝐿2 -energy 𝐶𝐸 [𝑢](𝑡) for the true trajectory and the surrogate prediction, showing monotone decay in both. Bottom: spatial mean 𝐶𝑀 [𝑢](𝑡), which the physics-conforming model preserves to near-machine precision. where 𝑞 denotes the displacement field and 𝑝 its velocity. In contrast to the heat equation, the wave equation is nondissipative: oscillatory … view at source ↗
Figure 10
Figure 10. Figure 10: Ablation study for the wave equation. Nor [PITH_FULL_IMAGE:figures/full_fig_p022_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Wave equation trajectory prediction for the displacement field 𝑞. Predicted fields (row 1), true fields (row 2), and absolute errors (row 3) are shown at eight time snapshots over 𝑡 ∈ [0, 5]. The physics-conforming Latent Twins captures the oscillatory wave dynamics with small localized errors. t= 0.63 t= 1.26 t= 1.88 t= 2.51 t= 3.14 t= 3.77 t= 4.39 t= 5.00 Predicted p True p Abs. error [PITH_FULL_IMAGE:… view at source ↗
Figure 12
Figure 12. Figure 12: Wave equation trajectory prediction for the velocity field 𝑝. Predicted fields (row 1), true fields (row 2), and absolute errors (row 3) are shown at eight time snapshots over 𝑡 ∈ [0, 5]. 5. Conclusions and outlook We introduced Physics-conforming Latent Twins, a latent solution-operator framework for learning surrogate dynamics that respect selected physical structures by design. The central idea is to u… view at source ↗
Figure 13
Figure 13. Figure 13: Conservation diagnostics for the wave Latent Twin. Left: normalized Hamiltonian 𝐶𝐻 (𝑡)∕𝐶𝐻 (0) for the true trajectory and surrogate prediction, with the predicted value deviating by approximately 2% from the reference over the full horizon. Right: relative latent norm variation 𝐶𝑍(𝑡)∕𝐶𝑍(0) − 1, confirming near-exact conservation of the quadratic latent invariant induced by the skew-symmetric latent operat… view at source ↗
read the original abstract

Surrogate models are central to scientific machine learning, where they enable fast prediction, simulation, inference, and control for complex physical systems. For time-dependent problems, however, accurate interpolation of training trajectories is not sufficient: reliable surrogates should also respect the conservation laws, invariants, admissibility conditions, and dissipative structures that give those trajectories physical meaning. We introduce Physics-conforming Latent Twins, a framework for learning latent surrogate solution operators whose dynamics satisfy selected physical principles by design. The method builds on the Latent Twin formulation by jointly learning an encoder, a decoder, and a latent flow map between arbitrary time-indexed states, while constraining the latent dynamics to preserve or dissipate prescribed structural quantities. We develop a constraint-transfer viewpoint that connects physical structure in the original state space with compatible constraints in latent space, and prove structure-preservation bounds showing how latent enforcement improves control of physical defects after decoding. We also derive algebraic conditions for latent flow maps that preserve linear and quadratic invariants or enforce dissipative inequalities. Numerical experiments on representative ODE and PDE benchmarks demonstrate improved constraint satisfaction, structural fidelity, and qualitative long-time behavior while maintaining accurate surrogate prediction.

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

1 major / 0 minor

Summary. The manuscript introduces Physics-conforming Latent Twins, a framework for learning latent surrogate solution operators for time-dependent physical systems. Building on the Latent Twin formulation, it jointly learns an encoder, decoder, and latent flow map while constraining the latent dynamics to preserve or dissipate prescribed structural quantities. The paper develops a constraint-transfer viewpoint connecting state-space physical structure to compatible latent constraints, proves structure-preservation bounds on how latent enforcement controls physical defects after decoding, derives algebraic conditions for latent flow maps that preserve linear/quadratic invariants or enforce dissipative inequalities, and reports numerical experiments on ODE and PDE benchmarks demonstrating improved constraint satisfaction, structural fidelity, and long-time behavior while maintaining surrogate accuracy.

Significance. If the claimed proofs and derivations hold, the work is significant for scientific machine learning because it supplies a systematic way to embed physical principles directly into latent-space dynamics, addressing the common failure of data-driven surrogates to respect conservation laws, invariants, and dissipation over long times. The structure-preservation bounds and algebraic conditions for invariant-preserving or dissipative latent maps constitute theoretical contributions that could guide the design of physics-conforming models. The numerical experiments are presented as evidence that accuracy can be retained alongside improved physical fidelity.

major comments (1)
  1. [Abstract] Abstract and constraint-transfer viewpoint: the central claims rest on the existence of compatible structural constraints that can be transferred from state space to latent space and jointly optimized with the encoder/decoder without destroying representational accuracy. The manuscript must supply the explicit construction of these transferred constraints, the precise statement of the structure-preservation bounds (including any assumptions on the decoder), and the full algebraic derivations for the invariant and dissipative conditions; without these details the load-bearing theoretical results cannot be verified.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the thorough review and for identifying the need to strengthen the presentation of the theoretical results. We address the major comment below and will incorporate the requested clarifications in a revised manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract and constraint-transfer viewpoint: the central claims rest on the existence of compatible structural constraints that can be transferred from state space to latent space and jointly optimized with the encoder/decoder without destroying representational accuracy. The manuscript must supply the explicit construction of these transferred constraints, the precise statement of the structure-preservation bounds (including any assumptions on the decoder), and the full algebraic derivations for the invariant and dissipative conditions; without these details the load-bearing theoretical results cannot be verified.

    Authors: We agree that the explicit constructions, precise bound statements, and full derivations are essential for verifiability. These elements are present in the manuscript (constraint transfer in Section 3 with the encoder-decoder commutativity construction; structure-preservation bounds in Theorem 4.1 under a Lipschitz decoder assumption; algebraic conditions for linear/quadratic invariants and dissipation in Sections 4.2–4.3), but we acknowledge they require expansion for clarity. In the revision we will add the explicit transferred-constraint formulas, restate the bounds with all decoder assumptions listed, and include the complete step-by-step algebraic derivations for the latent flow-map conditions. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation self-contained

full rationale

The abstract and reader's summary indicate that the constraint-transfer viewpoint, structure-preservation bounds, and algebraic conditions for invariant-preserving latent flow maps are presented as new derived contributions. The framework builds on a prior Latent Twin formulation via reference, but the core proofs and conditions do not reduce by construction to fitted parameters, self-defined quantities, or load-bearing self-citations whose validity depends on the present work. No quoted equations or steps exhibit the enumerated circular patterns such as self-definitional relations or renaming known results as novel predictions. The numerical experiments are described as demonstrating maintained accuracy alongside improved constraint satisfaction, consistent with independent validation rather than forced outcomes. This is the expected non-finding for a paper whose central claims retain independent mathematical content.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Review performed on abstract only; no explicit free parameters, axioms, or invented entities beyond the new framework itself are detailed in the provided text.

pith-pipeline@v0.9.1-grok · 5725 in / 1126 out tokens · 39760 ms · 2026-06-27T04:55:36.434346+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

79 extracted references · 7 canonical work pages · 2 internal anchors

  1. [1]

    Ascher , title =

    Uri M. Ascher , title =. Hybrid Systems: Computation and Control , series =. 2002 , doi =

    2002

  2. [2]

    Stewart , title =

    David E. Stewart , title =. 2011 , doi =

    2011

  3. [3]

    Ernst Hairer and Gerhard Wanner , title =

  4. [4]

    Computer Methods in Applied Mechanics and Engineering , volume=

    tLaSDI: Thermodynamics-informed latent space dynamics identification , author=. Computer Methods in Applied Mechanics and Engineering , volume=. 2024 , publisher=

    2024

  5. [5]

    Computer Methods in Applied Mechanics and Engineering , volume=

    Weak-form latent space dynamics identification , author=. Computer Methods in Applied Mechanics and Engineering , volume=. 2024 , publisher=

    2024

  6. [6]

    International Journal for Numerical Methods in Engineering , volume=

    Physics-Informed Active Learning With Simultaneous Weak-Form Latent Space Dynamics Identification , author=. International Journal for Numerical Methods in Engineering , volume=. 2025 , publisher=

    2025

  7. [7]

    arXiv preprint arXiv:2403.10748 , year=

    A comprehensive review of latent space dynamics identification algorithms for intrusive and non-intrusive reduced-order-modeling , author=. arXiv preprint arXiv:2403.10748 , year=

    work page arXiv

  8. [8]

    , series=

    Murray, James D. , series=. Mathematical Biology:. 2007 , publisher=

    2007

  9. [9]

    , title =

    Khalil, Hassan K. , title =

  10. [10]

    1990 , publisher =

    A Mathematical Introduction to Fluid Mechanics , author =. 1990 , publisher =

    1990

  11. [11]

    Advances in Neural Information Processing Systems , volume=

    Pinnacle: A comprehensive benchmark of physics-informed neural networks for solving pdes , author=. Advances in Neural Information Processing Systems , volume=

  12. [12]

    AI , volume=

    Understanding physics-informed neural networks: Techniques, applications, trends, and challenges , author=. AI , volume=. 2024 , publisher=

    2024

  13. [13]

    SIAM Journal on Scientific Computing , volume=

    Understanding and mitigating gradient flow pathologies in physics-informed neural networks , author=. SIAM Journal on Scientific Computing , volume=. 2021 , publisher=

    2021

  14. [14]

    Artificial Intelligence Review , volume=

    Physics-informed neural networks for PDE problems: A comprehensive review , author=. Artificial Intelligence Review , volume=. 2025 , publisher=

    2025

  15. [15]

    Applied Sciences , volume=

    Physics-informed neural networks: A review of methodological evolution, theoretical foundations, and interdisciplinary frontiers toward next-generation scientific computing , author=. Applied Sciences , volume=. 2025 , publisher=

    2025

  16. [16]

    Annual review of fluid mechanics , volume=

    The proper orthogonal decomposition in the analysis of turbulent flows , author=. Annual review of fluid mechanics , volume=. 1993 , publisher=

    1993

  17. [17]

    Journal of Fluid Mechanics , volume=

    Dynamic mode decomposition of numerical and experimental data , author=. Journal of Fluid Mechanics , volume=. 2010 , publisher=

    2010

  18. [18]

    Model Reduction and Approximation: Theory and Algorithms , volume=

    A tutorial introduction to the Loewner framework for model reduction , author=. Model Reduction and Approximation: Theory and Algorithms , volume=. 2017 , publisher=

    2017

  19. [19]

    2005 , publisher=

    Approximation of large-scale dynamical systems , author=. 2005 , publisher=

    2005

  20. [20]

    Journal of Computational Dynamics , volume=

    On dynamic mode decomposition: theory and applications , author=. Journal of Computational Dynamics , volume=

  21. [21]

    2016 , publisher=

    Dynamic Mode Decomposition: Data-Driven Modeling of Complex Systems , author=. 2016 , publisher=

    2016

  22. [22]

    arXiv preprint arXiv:2209.14855 , year=

    Continuous pde dynamics forecasting with implicit neural representations , author=. arXiv preprint arXiv:2209.14855 , year=

    work page arXiv

  23. [23]

    Advances in Neural Information Processing Systems , volume=

    Operator learning with neural fields: Tackling pdes on general geometries , author=. Advances in Neural Information Processing Systems , volume=

  24. [24]

    Computer Methods in Applied Mechanics and Engineering , volume=

    Lasdi: Parametric latent space dynamics identification , author=. Computer Methods in Applied Mechanics and Engineering , volume=. 2022 , publisher=

    2022

  25. [25]

    arXiv preprint arXiv:2206.02607 , year=

    CROM: Continuous reduced-order modeling of PDEs using implicit neural representations , author=. arXiv preprint arXiv:2206.02607 , year=

    work page arXiv

  26. [26]

    Journal of Computational Physics , volume=

    gLaSDI: Parametric physics-informed greedy latent space dynamics identification , author=. Journal of Computational Physics , volume=. 2023 , publisher=

    2023

  27. [27]

    Journal of Machine Learning Research , volume=

    Neural operator: Learning maps between function spaces with applications to pdes , author=. Journal of Machine Learning Research , volume=

  28. [28]

    Advances in neural information processing systems , volume=

    Implicit neural representations with periodic activation functions , author=. Advances in neural information processing systems , volume=

  29. [29]

    Physical Review E , volume=

    Using noisy or incomplete data to discover models of spatiotemporal dynamics , author=. Physical Review E , volume=. 2020 , publisher=

    2020

  30. [30]

    Nature communications , volume=

    Deep learning for universal linear embeddings of nonlinear dynamics , author=. Nature communications , volume=. 2018 , publisher=

    2018

  31. [31]

    Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences , volume=

    SINDy-PI: a robust algorithm for parallel implicit sparse identification of nonlinear dynamics , author=. Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences , volume=. 2020 , publisher=

    2020

  32. [32]

    Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences , volume=

    Learning partial differential equations via data discovery and sparse optimization , author=. Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences , volume=. 2017 , publisher=

    2017

  33. [33]

    Science advances , volume=

    Data-driven discovery of partial differential equations , author=. Science advances , volume=. 2017 , publisher=

    2017

  34. [34]

    Proceedings of the national academy of sciences , volume=

    Discovering governing equations from data by sparse identification of nonlinear dynamical systems , author=. Proceedings of the national academy of sciences , volume=. 2016 , publisher=

    2016

  35. [35]

    Journal of Machine Learning for Modeling and Computing , volume=

    Sparse L 1-Autoencoders for Scientific Data Compression , author=. Journal of Machine Learning for Modeling and Computing , volume=. 2025 , publisher=

    2025

  36. [36]

    Advances in neural information processing systems , volume=

    Learning Koopman invariant subspaces for dynamic mode decomposition , author=. Advances in neural information processing systems , volume=

  37. [37]

    Nonlinear Dynamics , volume=

    Spectral properties of dynamical systems, model reduction and decompositions , author=. Nonlinear Dynamics , volume=. 2005 , publisher=

    2005

  38. [38]

    Proceedings of the National Academy of Sciences , volume=

    Hamiltonian systems and transformation in Hilbert space , author=. Proceedings of the National Academy of Sciences , volume=

  39. [39]

    SIAM review , volume=

    A survey of projection-based model reduction methods for parametric dynamical systems , author=. SIAM review , volume=. 2015 , publisher=

    2015

  40. [40]

    2016 , publisher=

    Certified Reduced Basis Methods for Parametrized Partial Differential Equations , author=. 2016 , publisher=

    2016

  41. [41]

    2005 , publisher=

    Gradient flows: in metric spaces and in the space of probability measures , author=. 2005 , publisher=

    2005

  42. [42]

    2006 , publisher=

    Numerical optimization , author=. 2006 , publisher=

    2006

  43. [43]

    Ussr computational mathematics and mathematical physics , volume=

    Some methods of speeding up the convergence of iteration methods , author=. Ussr computational mathematics and mathematical physics , volume=. 1964 , publisher=

    1964

  44. [44]

    Nachrichten von der Gesellschaft der Wissenschaften zu G

    Noether, Emmy , title =. Nachrichten von der Gesellschaft der Wissenschaften zu G

  45. [45]

    Invariante

    Noether, Emmy , booktitle=. Invariante. 1983 , publisher=

    1983

  46. [46]

    Machine Learning: Science and Technology , volume=

    Latent Twins , author=. Machine Learning: Science and Technology , volume=. 2026 , publisher=

    2026

  47. [47]

    Nature machine intelligence , volume=

    Learning nonlinear operators via DeepONet based on the universal approximation theorem of operators , author=. Nature machine intelligence , volume=. 2021 , publisher=

    2021

  48. [48]

    Fourier Neural Operator for Parametric Partial Differential Equations

    Fourier neural operator for parametric partial differential equations , author=. arXiv preprint arXiv:2010.08895 , year=

    work page internal anchor Pith review Pith/arXiv arXiv 2010

  49. [49]

    Lecture Notes, Institute of Mathematics and Scientific Computing, University of Graz

    Model reduction using proper orthogonal decomposition , author=. Lecture Notes, Institute of Mathematics and Scientific Computing, University of Graz. see http://www. uni-graz. at/imawww/volkwein/POD. pdf , volume=

  50. [50]

    2021 , publisher=

    Classical Electrodynamics , author=. 2021 , publisher=

    2021

  51. [51]

    2006 , publisher=

    Mathematical Aspects of Classical and Celestial Mechanics , author=. 2006 , publisher=

    2006

  52. [52]

    1992 , publisher=

    Microeconomic Analysis , author=. 1992 , publisher=

    1992

  53. [53]

    1993 , publisher=

    Network Flows: Theory, Algorithms, and Applications , author=. 1993 , publisher=

    1993

  54. [54]

    1991 , publisher =

    Elements of Information Theory , author =. 1991 , publisher =

    1991

  55. [55]

    Journal of Computational physics , volume=

    Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations , author=. Journal of Computational physics , volume=. 2019 , publisher=

    2019

  56. [56]

    Nature Reviews Physics , volume=

    Physics-informed machine learning , author=. Nature Reviews Physics , volume=. 2021 , publisher=

    2021

  57. [57]

    Computer Methods in Applied Mechanics and Engineering , volume=

    When and why PINNs fail to train: A neural tangent kernel perspective , author=. Computer Methods in Applied Mechanics and Engineering , volume=. 2021 , publisher=

    2021

  58. [58]

    Advances in Neural Information Processing Systems (NeurIPS) , year=

    PINNacle: A Comprehensive Benchmark of Physics-Informed Neural Networks for Solving PDEs , author=. Advances in Neural Information Processing Systems (NeurIPS) , year=

  59. [59]

    International Conference on Learning Representations , year=

    HyperNetworks , author=. International Conference on Learning Representations , year=

  60. [60]

    Journal of Computational Physics , volume=

    Thermodynamically consistent physics-informed neural networks for hyperbolic systems , author=. Journal of Computational Physics , volume=. 2022 , publisher=

    2022

  61. [61]

    Advances in Neural Information Processing Systems , volume=

    Hamiltonian Neural Networks , author=. Advances in Neural Information Processing Systems , volume=

  62. [62]

    Proceedings of the Royal Society A , volume=

    Physics-informed dynamic mode decomposition , author=. Proceedings of the Royal Society A , volume=. 2023 , publisher=

    2023

  63. [63]

    Annual Conference on Neural Information Processing Systems , year=

    Guaranteeing Conservation Laws with Projection in Physics-Informed Neural Networks , author=. Annual Conference on Neural Information Processing Systems , year=

  64. [64]

    SIAM review , volume=

    The Mathematics of Infectious Diseases , author=. SIAM review , volume=. 2000 , publisher=

    2000

  65. [65]

    Campbell, S. L. and Gear, C. W. The index of general nonlinear D A E S. Numer. M ath. 1995

    1995

  66. [66]

    Slifka, M. K. and Whitton, J. L. Clinical implications of dysregulated cytokine production. J. M ol. M ed. 2000. doi:10.1007/s001090000086

    work page doi:10.1007/s001090000086 2000

  67. [67]

    Quasimonotonicity, regularity and duality for nonlinear systems of partial differential equations

    Hamburger, C. Quasimonotonicity, regularity and duality for nonlinear systems of partial differential equations. Ann. Mat. Pura. Appl. 1995

    1995

  68. [68]

    Geddes, K. O. and Czapor, S. R. and Labahn, G. Algorithms for C omputer A lgebra. 1992

    1992

  69. [69]

    Software engineering---from auxiliary to key technologies

    Broy, M. Software engineering---from auxiliary to key technologies. Software Pioneers. 1992

    1992

  70. [70]

    Conductive P olymers. 1981

    1981

  71. [71]

    Smith, S. E. Neuromuscular blocking drugs in man. Neuromuscular junction. H andbook of experimental pharmacology. 1976

    1976

  72. [72]

    Chung, S. T. and Morris, R. L. Isolation and characterization of plasmid deoxyribonucleic acid from Streptomyces fradiae. 1978

    1978

  73. [73]

    and AghaKouchak, A

    Hao, Z. and AghaKouchak, A. and Nakhjiri, N. and Farahmand, A. Global integrated drought monitoring and prediction system (GIDMaPS) data sets. 2014

    2014

  74. [74]

    Babichev, S. A. and Ries, J. and Lvovsky, A. I. Quantum scissors: teleportation of single-mode optical states by means of a nonlocal single photon. 2002

    2002

  75. [75]

    Wormholes in Maximal Supergravity

    Beneke, M. and Buchalla, G. and Dunietz, I. Mixing induced CP asymmetries in inclusive B decays. Phys. L ett. 1997. arXiv:0707.3168

    work page internal anchor Pith review Pith/arXiv arXiv 1997

  76. [76]

    deep SIP : deep learning of S upernova I a P arameters

    Stahl, B. deep SIP : deep learning of S upernova I a P arameters. 2020. ascl:2006.023

    2020

  77. [77]

    Abbott, T. M. C. and others. Dark Energy Survey Year 1 Results: Constraints on Extended Cosmological Models from Galaxy Clustering and Weak Lensing. Phys. Rev. D. 2019. doi:10.1103/PhysRevD.99.123505. arXiv:1810.02499

    work page doi:10.1103/physrevd.99.123505 2019

  78. [78]

    Physica D: Nonlinear Phenomena , volume=

    Model reduction for compressible flows using POD and Galerkin projection , author=. Physica D: Nonlinear Phenomena , volume=. 2004 , publisher=

    2004

  79. [79]

    Physical Review Research , volume=

    Dedalus: A flexible framework for numerical simulations with spectral methods , author=. Physical Review Research , volume=. 2020 , publisher=

    2020