pith. sign in

arxiv: 2606.09432 · v1 · pith:VN4PLDF2new · submitted 2026-06-08 · 💻 cs.LG

Graph Mamba Operator: A Latent Simulator for Interacting Particle Systems

Karn Tiwari , Niladri Dutta , N M Anoop Krishnan , Prathosh A P This is my paper

Pith reviewed 2026-06-27 17:05 UTC · model grok-4.3

classification 💻 cs.LG
keywords graph neural networksstate-space modelsinteracting particle systemslatent simulatorlong-horizon predictiondynamical systemsgraph mamba
0
0 comments X

The pith

GraMO couples graph interactions and temporal updates inside one linear recurrence with input-dependent coefficients to simulate interacting particle systems.

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

The paper presents GraMO as a latent-space simulator for dynamical systems such as N-body particles, motion capture, and robotics. It claims that standard graph neural networks accumulate errors over long horizons because they handle spatial interactions and temporal dynamics in separate stages or by sequencing nodes. GraMO instead merges graph-based interaction learning with state-space temporal updates into a single recurrence whose update remains linear in the latent state yet adapts via input-dependent coefficients. This design is said to capture multi-hop dependencies and global structure while delivering the lowest errors on the tested benchmarks and the biggest improvements when predictions extend far into the future.

Core claim

GraMO integrates state-space models with graph-based interaction learning by coupling graph-based interactions and temporal state updates within a single recurrence. The update is linear in the latent state, with input-dependent coefficients that adapt across regimes. On N-body systems, motion capture, and robotics datasets it records the lowest error across benchmarks and the largest gains in long-horizon prediction.

What carries the argument

The Graph Mamba Operator, which performs graph-based interactions and temporal state updates together inside one shared recurrence that stays linear in the latent state while using input-dependent coefficients.

If this is right

  • Lowest prediction error on N-body, motion capture, and robotics datasets compared with prior methods.
  • Largest accuracy gains appear precisely when the forecast horizon lengthens.
  • The single recurrence is claimed to handle multi-hop dependencies without extra mechanisms.
  • Input-dependent coefficients allow the same linear update to adapt across different dynamical regimes.

Where Pith is reading between the lines

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

  • The approach could be tested on additional domains such as fluid or molecular dynamics where both spatial topology and long time scales matter.
  • If the linear recurrence with input-dependent coefficients suffices, similar single-stage designs might replace staged spatial-temporal modules in other sequence models.
  • Success on particle systems suggests the method may scale to larger numbers of particles provided the graph construction remains efficient.

Load-bearing premise

That combining graph interactions and temporal updates inside one shared recurrence is enough to capture multi-hop dependencies and global structure without creating new error modes.

What would settle it

If a model that keeps spatial and temporal stages separate matches or beats GraMO on long-horizon prediction error for the N-body benchmark, the advantage of the single-recurrence coupling would be called into question.

Figures

Figures reproduced from arXiv: 2606.09432 by Karn Tiwari, Niladri Dutta, N M Anoop Krishnan, Prathosh A P.

Figure 1
Figure 1. Figure 1: Overview. The model maps past graph sequences {Gt} T t=0 to future trajectories {Gt} T +∆T t=T +1 using L stacked GraMO blocks. (Left) Overall pipeline. (Right) A single GraMO layer showing the discretized update and gated skip connection (see Appendix Algorithm 1). Discrete-Time Formulation. Under a discrete-time setting with step size ∆, the discrete dynamics can be written as follows: z[k] = Kz ¯ [k − 1… view at source ↗
Figure 2
Figure 2. Figure 2: Efficiency and Visualization. (Left) Training time (ms) versus FMSE (×10−1 ) for baselines on MoCap (Walk) dataset. (Right) MoCap (Run) trajectory visualization, where predicted trajectories are shown with a Blue color gradient over time, and the ground-truth final state is shown in Green. Visual Demonstrations [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Visualization of Per-step trajectory error as a function of prediction horizon for [PITH_FULL_IMAGE:figures/full_fig_p028_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Visualization of trajectories generated by [PITH_FULL_IMAGE:figures/full_fig_p028_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Visualization of trajectories generated by [PITH_FULL_IMAGE:figures/full_fig_p029_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Visualization of trajectories generated by [PITH_FULL_IMAGE:figures/full_fig_p029_6.png] view at source ↗
read the original abstract

Modeling interacting dynamical systems requires capturing spatial interactions alongside long-range temporal dependencies. Graph neural networks (GNNs) provide a natural representation but typically rely on autoregressive rollouts and treat spatial and temporal dynamics separately, leading to error accumulation over long horizons. Existing approaches also focus on local interactions and short temporal contexts, limiting their ability to capture multi-hop dependencies and global structure. We introduce the Graph Mamba Operator (GraMO), a latent-space simulator that integrates state-space models with graph-based interaction learning. In contrast to prior work that sequences nodes or applies spatial and temporal updates in separate stages, GraMO couples graph-based interactions and temporal state updates within a single recurrence. The update is linear in the latent state, with input-dependent coefficients that adapt across regimes. We evaluate GraMO on N-body systems, motion capture, and robotics datasets, achieving the lowest error across benchmarks and the largest gains in long-horizon 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 paper introduces the Graph Mamba Operator (GraMO), a latent-space simulator for interacting dynamical systems that integrates state-space models with graph-based interaction learning. Unlike prior work that sequences nodes or separates spatial and temporal updates, GraMO couples graph-based interactions and temporal state updates inside a single recurrence that remains linear in the latent state while using input-dependent coefficients. The method is evaluated on N-body systems, motion capture, and robotics datasets and is claimed to achieve the lowest error across benchmarks with the largest gains on long-horizon prediction.

Significance. If the central construction holds, GraMO would offer a unified recurrence that avoids error accumulation from separate spatial/temporal stages and autoregressive rollouts while adapting across regimes via input-dependent coefficients. The multi-dataset evaluation and emphasis on long-horizon gains constitute a concrete strength. However, the significance depends on whether the single-recurrence design actually transmits multi-hop and global information without reintroducing the limitations the abstract attributes to prior methods.

major comments (1)
  1. [Abstract] Abstract: the central claim that embedding graph interactions inside one shared SSM recurrence (linear in latent state, input-dependent coefficients) suffices to capture multi-hop dependencies and global structure is load-bearing yet unsupported by any described mechanism. No mention is made of multi-layer propagation, global attention, explicit hop counting, or analysis showing that information travels beyond immediate neighbors while preserving the linear SSM properties; if the graph operator remains local per time step, the architecture risks reintroducing the error accumulation and limited context the paper seeks to solve.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback on the abstract. We address the concern about the mechanism for multi-hop and global dependencies below, clarifying the construction from the full manuscript while agreeing that the abstract requires expansion for clarity.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that embedding graph interactions inside one shared SSM recurrence (linear in latent state, input-dependent coefficients) suffices to capture multi-hop dependencies and global structure is load-bearing yet unsupported by any described mechanism. No mention is made of multi-layer propagation, global attention, explicit hop counting, or analysis showing that information travels beyond immediate neighbors while preserving the linear SSM properties; if the graph operator remains local per time step, the architecture risks reintroducing the error accumulation and limited context the paper seeks to solve.

    Authors: Section 3 of the manuscript defines GraMO by embedding the graph interaction directly into the input-dependent coefficients of a single linear SSM recurrence (specifically modulating the state transition and projection terms using both node features and adjacency). Local neighbor information is thus incorporated at every time step, while the recurrent dynamics over the horizon enable multi-hop propagation through successive updates without separate spatial stages or autoregressive rollouts. The design avoids explicit multi-layer GNN propagation or attention by relying on this unified recurrence, which remains linear in the latent state. We agree the abstract is too concise on this point and will revise it to reference the coefficient modulation mechanism and its implications for information flow. revision: yes

Circularity Check

0 steps flagged

No circularity detected; architecture claims are independent of inputs

full rationale

The paper introduces GraMO as a new latent simulator coupling graph interactions and SSM-style temporal updates inside one recurrence, with claims resting on empirical results across N-body, motion capture, and robotics benchmarks. No equations, fitted parameters, or self-citations are presented that reduce any prediction or uniqueness claim to a quantity defined inside the paper by construction. The central modeling choice is presented as a design decision rather than a derived necessity, and performance gains are evaluated externally, making the derivation self-contained.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only; the central claim rests on the unstated assumption that standard graph and SSM building blocks can be fused without additional axioms. No free parameters or invented entities are visible.

pith-pipeline@v0.9.1-grok · 5698 in / 1173 out tokens · 13095 ms · 2026-06-27T17:05:34.263187+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

155 extracted references · 1 canonical work pages

  1. [1]

    Scaling Learning Algorithms Towards

    Bengio, Yoshua and LeCun, Yann , booktitle =. Scaling Learning Algorithms Towards

  2. [2]

    and Osindero, Simon and Teh, Yee Whye , journal =

    Hinton, Geoffrey E. and Osindero, Simon and Teh, Yee Whye , journal =. A Fast Learning Algorithm for Deep Belief Nets , volume =

  3. [3]

    2016 , publisher=

    Deep learning , author=. 2016 , publisher=

    2016

  4. [4]

    arXiv preprint arXiv:2312.00752 , year=

    Mamba: Linear-time sequence modeling with selective state spaces , author=. arXiv preprint arXiv:2312.00752 , year=

    Pith/arXiv arXiv

  5. [5]

    International conference on machine learning , pages=

    Neural relational inference for interacting systems , author=. International conference on machine learning , pages=. 2018 , organization=

    2018

  6. [6]

    International conference on machine learning , pages=

    E (n) equivariant graph neural networks , author=. International conference on machine learning , pages=. 2021 , organization=

    2021

  7. [7]

    Science advances , volume=

    Machine learning of accurate energy-conserving molecular force fields , author=. Science advances , volume=. 2017 , publisher=

    2017

  8. [8]

    International Conference on Learning Representations , year=

    Spherical Message Passing for 3D Molecular Graphs , author=. International Conference on Learning Representations , year=

  9. [9]

    The Twelfth International Conference on Learning Representations , year=

    Improved Techniques for Training Consistency Models , author=. The Twelfth International Conference on Learning Representations , year=

  10. [10]

    Advances in Neural Information Processing Systems , volume=

    Learning physical dynamics with subequivariant graph neural networks , author=. Advances in Neural Information Processing Systems , volume=

  11. [11]

    International Conference on Machine Learning , pages=

    Consistency Models , author=. International Conference on Machine Learning , pages=. 2023 , organization=

    2023

  12. [12]

    International Conference on Learning Representations , year=

    Semi-Supervised Classification with Graph Convolutional Networks , author=. International Conference on Learning Representations , year=

  13. [13]

    Worrall and Max Welling , booktitle=

    Johannes Brandstetter and Daniel E. Worrall and Max Welling , booktitle=. Message Passing Neural. 2022 , url=

    2022

  14. [14]

    Advances in neural information processing systems , volume=

    Attention is all you need , author=. Advances in neural information processing systems , volume=

  15. [15]

    arXiv preprint arXiv:2305.18046 , year=

    Implicit Transfer Operator Learning: Multiple Time-Resolution Surrogates for Molecular Dynamics , author=. arXiv preprint arXiv:2305.18046 , year=

    arXiv

  16. [16]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    Deep residual learning for image recognition , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  17. [17]

    International Conference on Machine Learning , pages=

    Geometric latent diffusion models for 3d molecule generation , author=. International Conference on Machine Learning , pages=. 2023 , organization=

    2023

  18. [18]

    1977 , publisher=

    Linear representations of finite groups , author=. 1977 , publisher=

    1977

  19. [19]

    International Conference on Machine Learning , pages=

    Fast sampling of diffusion models via operator learning , author=. International Conference on Machine Learning , pages=. 2023 , organization=

    2023

  20. [20]

    International Conference on Learning Representations , year=

    Fourier Neural Operator for Parametric Partial Differential Equations , author=. International Conference on Learning Representations , year=

  21. [21]

    Cohen and Mario Geiger and Jonas Köhler and Max Welling , booktitle=

    Taco S. Cohen and Mario Geiger and Jonas Köhler and Max Welling , booktitle=. Spherical. 2018 , url=

    2018

  22. [22]

    The Eleventh International Conference on Learning Representations , year=

    Spatial Attention Kinetic Networks with E(n)-Equivariance , author=. The Eleventh International Conference on Learning Representations , year=

  23. [23]

    International Conference on Learning Representations , year=

    GeoDiff: A Geometric Diffusion Model for Molecular Conformation Generation , author=. International Conference on Learning Representations , year=

  24. [24]

    arXiv preprint arXiv:2003.03485 , year=

    Neural operator: Graph kernel network for partial differential equations , author=. arXiv preprint arXiv:2003.03485 , year=

    Pith/arXiv arXiv 2003

  25. [25]

    arXiv preprint arXiv:2210.17051 , year=

    Accelerating Carbon Capture and Storage Modeling using Fourier Neural Operators , author=. arXiv preprint arXiv:2210.17051 , year=

    arXiv

  26. [26]

    The Seismic Record , volume=

    Seismic wave propagation and inversion with Neural Operators , author=. The Seismic Record , volume=. 2021 , publisher=

    2021

  27. [27]

    AAAI Workshop on Deep Learning on Graphs: Methods and Applications , year=

    Geometry-Complete Perceptron Networks for 3D Molecular Graphs , author=. AAAI Workshop on Deep Learning on Graphs: Methods and Applications , year=

  28. [28]

    2022 , editor =

    Du, Weitao and Zhang, He and Du, Yuanqi and Meng, Qi and Chen, Wei and Zheng, Nanning and Shao, Bin and Liu, Tie-Yan , booktitle =. 2022 , editor =

    2022

  29. [29]

    Journal of Machine Learning Research , volume=

    On universal approximation and error bounds for Fourier Neural Operators , author=. Journal of Machine Learning Research , volume=. 2021 , publisher=

    2021

  30. [30]

    arXiv preprint arXiv:2108.08481 , year=

    Neural operator: Learning maps between function spaces , author=. arXiv preprint arXiv:2108.08481 , year=

    arXiv

  31. [31]

    Advances in Neural Information Processing Systems , volume=

    Equivariant graph hierarchy-based neural networks , author=. Advances in Neural Information Processing Systems , volume=

  32. [32]

    arXiv preprint arXiv:2007.04954 , year=

    Threedworld: A platform for interactive multi-modal physical simulation , author=. arXiv preprint arXiv:2007.04954 , year=

    arXiv 2007

  33. [33]

    URL: https://figshare

    Molecular dynamics trajectory for benchmarking MDAnalysis , author=. URL: https://figshare. com/articles/Molecular\_dynamics\_ trajectory\_for\_benchmarking\_MDAnalysis/5108170, doi , volume=

    arXiv

  34. [34]

    arXiv preprint arXiv:2202.07230 , year=

    Geometrically equivariant graph neural networks: A survey , author=. arXiv preprint arXiv:2202.07230 , year=

    arXiv

  35. [35]

    International Conference on Learning Representations , year=

    Graph Information Bottleneck for Subgraph Recognition , author=. International Conference on Learning Representations , year=

  36. [36]

    2021 , journal=

    E(n) Equivariant Graph Neural Networks , author=. 2021 , journal=. 2102.09844 , archivePrefix=

    arXiv 2021

  37. [37]

    science , volume=

    How to grow a mind: Statistics, structure, and abstraction , author=. science , volume=. 2011 , publisher=

    2011

  38. [38]

    arXiv preprint arXiv:1802.08219 , year=

    Tensor field networks: Rotation-and translation-equivariant neural networks for 3d point clouds , author=. arXiv preprint arXiv:1802.08219 , year=

    Pith/arXiv arXiv

  39. [39]

    arXiv preprint arXiv:1910.00753 , year=

    Equivariant flows: sampling configurations for multi-body systems with symmetric energies , author=. arXiv preprint arXiv:1910.00753 , year=

    arXiv 1910

  40. [40]

    Hamiltonian Neural Networks , url =

    Greydanus, Samuel and Dzamba, Misko and Yosinski, Jason , booktitle =. Hamiltonian Neural Networks , url =

  41. [41]

    Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , month =

    Xing, Yifan and He, Tong and Xiao, Tianjun and Wang, Yongxin and Xiong, Yuanjun and Xia, Wei and Wipf, David and Zhang, Zheng and Soatto, Stefano , title =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , month =. 2021 , pages =

    2021

  42. [42]

    Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence, (IJCAI) , year =

    Hierarchical Graph Convolutional Networks for Semi-supervised Node Classification , author =. Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence, (IJCAI) , year =

  43. [43]

    Hierarchical Graph Representation Learning with Differentiable Pooling , url =

    Ying, Zhitao and You, Jiaxuan and Morris, Christopher and Ren, Xiang and Hamilton, Will and Leskovec, Jure , booktitle =. Hierarchical Graph Representation Learning with Differentiable Pooling , url =

  44. [44]

    Proceedings of the 36th International Conference on Machine Learning , year =

    Self-Attention Graph Pooling , author =. Proceedings of the 36th International Conference on Machine Learning , year =

  45. [45]

    2018 , eprint=

    Towards Sparse Hierarchical Graph Classifiers , author=. 2018 , eprint=

    2018

  46. [46]

    International Conference on Learning Representations , year=

    GraphZoom: A Multi-level Spectral Approach for Accurate and Scalable Graph Embedding , author=. International Conference on Learning Representations , year=

  47. [47]

    ICML Graph Representation Learning and Beyond (GRL+) Workhop , year=

    Hierarchical Inter-Message Passing for Learning on Molecular Graphs , author=. ICML Graph Representation Learning and Beyond (GRL+) Workhop , year=

  48. [48]

    and Guan, Yuqiang and Kulis, Brian , journal=

    Dhillon, Inderjit S. and Guan, Yuqiang and Kulis, Brian , journal=. Weighted Graph Cuts without Eigenvectors A Multilevel Approach , year=

  49. [49]

    International Conference on Medical image computing and computer-assisted intervention , pages=

    U-net: Convolutional networks for biomedical image segmentation , author=. International Conference on Medical image computing and computer-assisted intervention , pages=. 2015 , organization=

    2015

  50. [50]

    arXiv preprint arXiv:2110.03336 , year=

    Frame averaging for invariant and equivariant network design , author=. arXiv preprint arXiv:2110.03336 , year=

    arXiv

  51. [51]

    ICLR 2022 Workshop on Geometrical and Topological Representation Learning , year=

    Sign and Basis Invariant Networks for Spectral Graph Representation Learning , author=. ICLR 2022 Workshop on Geometrical and Topological Representation Learning , year=

    2022

  52. [52]

    International Conference on Learning Representations , year=

    Equivariant Graph Mechanics Networks with Constraints , author=. International Conference on Learning Representations , year=

  53. [53]

    arXiv preprint arXiv:2006.10503 , year=

    Se (3)-transformers: 3d roto-translation equivariant attention networks , author=. arXiv preprint arXiv:2006.10503 , year=

    arXiv 2006

  54. [54]

    arXiv preprint arXiv:1612.00222 , year=

    Interaction networks for learning about objects, relations and physics , author=. arXiv preprint arXiv:1612.00222 , year=

    Pith/arXiv arXiv

  55. [55]

    IEEE transactions on neural networks and learning systems , volume=

    A comprehensive survey on graph neural networks , author=. IEEE transactions on neural networks and learning systems , volume=. 2020 , publisher=

    2020

  56. [56]

    AAAI , year=

    Scalable Graph Networks for Particle Simulations , author=. AAAI , year=

  57. [57]

    Mathematics of control, signals and systems , volume=

    Approximation by superpositions of a sigmoidal function , author=. Mathematics of control, signals and systems , volume=. 1989 , publisher=

    1989

  58. [58]

    Neural networks , volume=

    Approximation capabilities of multilayer feedforward networks , author=. Neural networks , volume=. 1991 , publisher=

    1991

  59. [59]

    arXiv preprint arXiv:1806.08047 , year=

    Flexible neural representation for physics prediction , author=. arXiv preprint arXiv:1806.08047 , year=

    Pith/arXiv arXiv

  60. [60]

    arXiv preprint arXiv:1909.12790 , year=

    Hamiltonian graph networks with ode integrators , author=. arXiv preprint arXiv:1909.12790 , year=

    arXiv 1909

  61. [61]

    International Conference on Machine Learning , pages=

    Generalizing convolutional neural networks for equivariance to lie groups on arbitrary continuous data , author=. International Conference on Machine Learning , pages=. 2020 , organization=

    2020

  62. [62]

    arXiv preprint arXiv:1412.6980 , year=

    Adam: A method for stochastic optimization , author=. arXiv preprint arXiv:1412.6980 , year=

    Pith/arXiv arXiv

  63. [63]

    International Conference on Machine Learning , pages=

    Lietransformer: equivariant self-attention for lie groups , author=. International Conference on Machine Learning , pages=. 2021 , organization=

    2021

  64. [64]

    arXiv preprint arXiv:2006.12745 , year=

    Learning physical constraints with neural projections , author=. arXiv preprint arXiv:2006.12745 , year=

    arXiv 2006

  65. [65]

    arXiv preprint arXiv:2010.13581 , year=

    Simplifying hamiltonian and lagrangian neural networks via explicit constraints , author=. arXiv preprint arXiv:2010.13581 , year=

    arXiv 2010

  66. [66]

    Praktische Verfahren der Gleichungsaufl

    Mises, RV and Pollaczek-Geiringer, Hilda , journal=. Praktische Verfahren der Gleichungsaufl. 1929 , publisher=

    1929

  67. [67]

    Biopolymers: original Research on biomolecules , volume=

    Development and current status of the CHARMM force field for nucleic acids , author=. Biopolymers: original Research on biomolecules , volume=. 2000 , publisher=

    2000

  68. [68]

    The Journal of Chemical Physics , volume=

    Schnet--a deep learning architecture for molecules and materials , author=. The Journal of Chemical Physics , volume=. 2018 , publisher=

    2018

  69. [69]

    arXiv preprint arXiv:1906.04015 , year=

    Cormorant: Covariant molecular neural networks , author=. arXiv preprint arXiv:1906.04015 , year=

    arXiv 1906

  70. [70]

    arXiv preprint arXiv:2105.03902 , year=

    Learning gradient fields for molecular conformation generation , author=. arXiv preprint arXiv:2105.03902 , year=

    arXiv

  71. [71]

    Carnegie-Mellon Motion Capture Database , author=

  72. [72]

    International Conference on Learning Representations , year=

    Lagrangian fluid simulation with continuous convolutions , author=. International Conference on Learning Representations , year=

  73. [73]

    Proceedings of the 36th International Conference on Machine Learning , pages =

    Graph U-Nets , author =. Proceedings of the 36th International Conference on Machine Learning , pages =. 2019 , editor =

    2019

  74. [74]

    International Conference on Learning Representations , year=

    Geometric and Physical Quantities improve E(3) Equivariant Message Passing , author=. International Conference on Learning Representations , year=

  75. [75]

    PointConv: Deep Convolutional Networks on 3D Point Clouds , publisher =

    Wu, Wenxuan and Qi, Zhongang and Fuxin, Li , keywords =. PointConv: Deep Convolutional Networks on 3D Point Clouds , publisher =. 2018 , copyright =. doi:10.48550/ARXIV.1811.07246 , url =

    work page doi:10.48550/arxiv.1811.07246 2018

  76. [76]

    International Conference on Machine Learning , pages=

    Learning to simulate complex physics with graph networks , author=. International Conference on Machine Learning , pages=. 2020 , organization=

    2020

  77. [77]

    arXiv preprint arXiv:2010.03409 , year=

    Learning mesh-based simulation with graph networks , author=. arXiv preprint arXiv:2010.03409 , year=

    arXiv 2010

  78. [78]

    arXiv preprint arXiv:1907.04490 , year=

    Deep lagrangian networks: Using physics as model prior for deep learning , author=. arXiv preprint arXiv:1907.04490 , year=

    Pith/arXiv arXiv 1907

  79. [79]

    arXiv preprint arXiv:1810.01566 , year=

    Learning particle dynamics for manipulating rigid bodies, deformable objects, and fluids , author=. arXiv preprint arXiv:1810.01566 , year=

    Pith/arXiv arXiv

  80. [80]

    International conference on machine learning , pages=

    Neural message passing for quantum chemistry , author=. International conference on machine learning , pages=. 2017 , organization=

    2017

Showing first 80 references.