Pith. sign in

REVIEW 3 major objections 5 minor 56 references

AdS-GNN -- a Conformally Equivariant Graph Neural Network

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A graph neural network that lifts point clouds into Anti-de Sitter space achieves approximate equivariance under the full conformal group, and on the 2D Ising model it recovers the conformal dimensions $\Delta_\sigma=1/8$ and…

desk verdict Novel AdS lift plus proper-distance message passing gives an interesting scale- and rotation-equivariant GNN, but the headline full conformal equivariance is not delivered and the new SCT capability is tested only once on MNIST. read the letter →

arxiv 2505.12880 v1 pith:2O3SGPWU submitted 2025-05-19 cs.LG cs.AIhep-th

Maksim Zhdanov , Nabil Iqbal , Erik Bekkers , Patrick Forré This is my paper · ORCID
classification cs.LGcs.AIhep-th
keywords conformalequivariancegraphneuralnetworksAnti-deSitterspacepointcloudsscaleinvarianceIsingmodelfieldtheorydimensions
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 tries to build a graph neural network whose predictions are unchanged when the input point cloud is subjected to any conformal (angle-preserving) transformation of $\mathbb{R}^d$, including scalings and special conformal transformations that other equivariant architectures do not handle. The route is to lift each point cloud into $\mathrm{AdS}_{d+1}$, where conformal transformations of the boundary act as ordinary isometries, and then run message passing keyed on the AdS proper distance. The paper argues that this makes the network exactly equivariant once the data is lifted, with only a mild symmetry breaking coming from the lift itself. On the 2D Ising conformal field theory, the trained network predicts $N$-point correlation functions and its readout exponents reproduce the known scaling dimensions $\Delta_\sigma=1/8$ and $\Delta_\epsilon=1$.

What carries the argument

The load-bearing object is the AdS lift: Algorithm 1 embeds every point at boundary scale $z_0$, computes the AdS center of mass of its $k_{\mathrm{lift}}$ nearest neighbours using the Galperin formula, and re-embeds the point at the resulting height $\hat{z}_i$, so that $X_i=(x_i,\hat{z}_i)$ lives in $\mathrm{AdS}_{d+1}$. This turns each conformal transformation of the boundary into an isometry of the bulk, and the invariant proper distance $\cosh D(X_i,X_j)=(z_i^2+z_j^2+\|x_i-x_j\|^2)/(2z_iz_j)$ becomes the only geometric input to the message function. The same construction supplies the feature lift $h_i^{\mathrm{lifted}}=\hat{z}_i^{\Delta} h_i^{\mathrm{input}}$ and the conjugate readout $O(x_i)=\hat{z}_i^{-\Delta} h_i^{\mathrm{final}}$, which is what lets the network learn and expose conformal dimensions.

What would settle it

Test the full pipeline under a family of special conformal transformations on point clouds with sharply varying local density, or inspect Algorithm 1 directly on a pair of coincident points: if the equivariance error grows with density contrast, or if the computed $\hat{z}_i$ diverges or vanishes as $z_0\to0$, the central claim of mild, controlled symmetry breaking would be falsified.

Watch

Extended reading notes

Core claim

The central discovery is a construction that reduces conformal equivariance for point-cloud networks to isometric equivariance: each point $x_i$ is embedded at a scale-dependent height $z_i$ in $\mathrm{AdS}_{d+1}$, computed with a nearest-neighbour center of mass so that $z_i$ encodes the local length scale of the point cloud. Once the lifted points $X_i=(x_i,z_i)$ are fixed, the message-passing layer conditioned on the AdS proper distance $D(X_i,X_j)$ is exactly invariant under the isometry group of $\mathrm{AdS}_{d+1}$, which is the same group as the global conformal group $\mathrm{PO}(d+1,1)$ of $\mathbb{R}^d$. The paper reports that the residual symmetry breaking from the lift is small, and that on the 2D Ising model the readout $\log \mathrm{Pred}_a = \mathrm{AdSGNN}_a - \Delta_a \sum_i \log \hat{z}_i$ turns the trained network into a calculator for conformal dimensions, recovering $\Delta_\sigma=1/8$ and $\Delta_\epsilon=1$.

Load-bearing premise

The construction stands on the assumption that the nearest-neighbour center-of-mass lift assigns every point a finite, non-degenerate height $z_i$ in the $z_0\to0$ limit and that the resulting breaking of special conformal invariance is small enough not to destroy the equivariance benefit.

Editorial extensions

If this is right

  • A network built this way is exactly invariant under the isometry group of $\mathrm{AdS}_{d+1}$, so once a point cloud is lifted its output is conformally invariant by construction, independent of how the input was discretised.
  • Because the height $\hat{z}_i$ tracks local density, the model should be scale-invariant and should keep its predictions accurate on inputs whose spatial extent lies far outside the training range.
  • The conformal-dimension readout in the Ising experiments gives a direct, interpretable interface between a trained network and the universal data of a conformal field theory.
  • The same architecture transfers across system sizes: a network trained on correlation functions of $N=8$ points can predict $N=16$ correlation functions better than the flat-space baselines, suggesting it has learned the physics rather than a lookup table.

Reading between the lines

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

  • Beyond the paper, the same readout trick could be aimed at the three-point coefficients $c_{abc}$, which together with the conformal dimensions determine all higher correlation functions in a conformal field theory; the paper raises this question but does not answer it.
  • The tested special-conformal perturbation is a single one-parameter family; a natural stress test would apply a wider range of such transformations to point clouds with strongly nonuniform density, where the center-of-mass lift's symmetry breaking should be largest.
  • The scalar-feature restriction points to a plausible extension: combining the AdS lift with orientation-carrying feature spaces would give vector-valued conformal primaries and could close the expressivity gap the authors observe on rotation-sensitive tasks.
  • If the claimed finite limit of the lift as $z_0\to0$ can be established rigorously, the construction would apply to degenerate configurations with coincident or nearly coincident points, where the current empirical check is weakest.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper introduces AdS-GNN, a graph neural network for point clouds that lifts input points from Euclidean space R^d to Anti de Sitter space AdS_{d+1} using a per-point scale coordinate derived from the local neighborhood via an AdS center-of-mass computation (Algorithm 1). Message passing is then performed on the lifted points using the AdS proper distance, which is invariant under the isometry group PO(d+1,1) of AdS, the same group as the global conformal group of R^d. The authors state that the resulting network is equivariant under general conformal transformations, with exact invariance under translations, rotations, and scalings, and only mild breaking under special conformal transformations (SCTs) introduced by the lift. They validate the model on superpixel MNIST, shape segmentation, PascalVOC, and on predicting N-point correlation functions of the 2D Ising CFT, where the learned conformal dimensions are close to the exact values Delta_sigma=1/8 and Delta_epsilon=1. A substantial appendix develops the conformal geometry background and details of the AdS construction.

Significance. If the approximate conformal equivariance claim is accepted, the paper makes a useful contribution: it provides a simple, computationally cheap message-passing architecture that is exactly isometry-equivariant on AdS and that demonstrably generalizes under scaling and across system sizes in the Ising task. The clean two-point-function argument and the interpretable extraction of conformal dimensions are attractive and go beyond standard benchmark reporting. The mathematical appendix is self-contained and grounds the construction in standard results on the conformal group and AdS geometry. However, the central novelty advertised in the abstract and introduction—equivariance under special conformal transformations—is not as strongly supported as the rest of the paper. The Ising experiments, which are the most physically meaningful evidence, test scale covariance rather than SCT equivariance, and the only SCT experiment is a single MNIST augmentation test with incomplete reporting. The paper is therefore of genuine interest but needs substantial qualification and additional evidence before the central claim can be accepted at face value.

major comments (3)
  1. [Abstract, Section 1, Section 4.1, Section 6.1] The paper's headline claim of equivariance under general conformal transformations is stronger than what is actually constructed. Algorithm 1 and the readout in Eqs. (12), (15), and (17) are exactly equivariant only under translations, rotations, and scalings; the lift is explicitly stated to break special conformal transformations, and Section 6.1 concedes this. The abstract and the sentences in Sections 1 and 4.2 that describe a 'conformal group equivariant GNN' should therefore be qualified as approximate equivariance under the full conformal group. This is not merely a wording issue: SCT equivariance is the unique new capability of the model relative to scale-equivariant networks, so the present formulation overstates the contribution.
  2. [Section 5.1, Figure 4, Section 5.2] The empirical evidence that the SCT breaking is 'mild' is insufficient for the weight placed on it. Figure 4 reports a single MNIST experiment with one one-parameter family b=(0,b2) of special conformal transformations, and the text does not state the values of b used, the error magnitude, or the hyperparameters z0, k_lift, and k_con for that run. More importantly, the Ising experiments in Section 5.2, where the physics claims are made, never apply SCTs to the point sets; they test scale covariance, which is already enforced by the readout structure of Eq. (17). Please add quantitative SCT equivariance error measurements on the point-cloud tasks (including the Ising correlation task), or explicitly restrict the claims to approximate scale-and-rotation equivariance with an empirical statement of the SCT error.
  3. [Section 4.1, Algorithm 1, Section B.3] The claim that the AdS center-of-mass coordinate \hat z_i has a finite, scale-covariant limit as z0->0 is asserted without proof and depends on the KNN convention in Algorithm 1. For the value k_lift=1 used in all Ising experiments (Section B.3), the behavior is ambiguous: if KNN includes the point itself, then \hat z_i = z0 and the z-coordinate does not encode any local scale; if KNN excludes the point itself, this convention must be stated. Please specify the KNN convention, provide a derivation or argument for the finite limit for general k_lift, and report the resulting range of \hat z_i values. This matters because the scale-covariance of the readout in Eqs. (12), (15), and (17) relies on \hat z_i transforming as z -> lambda z under a global scaling.
minor comments (5)
  1. [Section 3.2] The text refers to 'Figure ??' when describing the two connected components of the hyperboloid; this is a missing reference and should be fixed.
  2. [Section B.2] There is a typo in 'how a a model trained on a given number of nodes' — the duplicated article should be removed.
  3. [Figure 4] The figure would be much more informative with labeled axes, the range of the SCT parameter b, and error bars or the number of trials; please add these details.
  4. [Section 4.2] The sentence 'which yields an efficient conformal group equivariant GNN' should be adjusted to 'approximately conformally equivariant GNN' to match the acknowledged SCT breaking in the lift.
  5. [Eq. (17)] The expression log(\hat z_i) uses a dimensionful quantity; if a reference scale or normalization is intended, it should be stated explicitly, especially because the logarithmic term is central to the learned conformal dimensions.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular reduction; central equivariance construction is self-contained, though the advertised full conformal equivariance is conceded and only partially verified.

full rationale

The paper's derivation chain is not circular. Equivariance after lifting is established by the explicit PO(d+1,1)-invariant proper distance in Eq. (9) and the message function in Eq. (14); the AdS lift in Algorithm 1 is explicitly stated to be equivariant only under rotations, translations, and scaling and to 'gently break' special conformal transformations (Sec. 4.1), a concession repeated in Sec. 6.1. The SCT generalization is therefore an empirical claim tested on MNIST (Fig. 4), not a consequence smuggled in by definition or by self-citation. The learned conformal dimensions in Table 2 are trainable readout parameters in Eq. (17) fitted to the same Ising correlation functions, so reading them off is supervised interpretability, not a held-out prediction forced by construction. Citations [11,13] are by the same group but are used only as background for isometry-equivariant networks; the core invariance of the proper distance is elementary and developed self-containedly in Eqs. (6)-(9). The main weakness is an overstated abstract claim versus the acknowledged SCT breaking, which is a correctness and evidence issue, not circularity.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The network contributes a new geometric construction but does not postulate new physics entities. The learned Delta values are fitted parameters, not built-in constraints. The main unproved input is the well-definedness and physical relevance of the z-lift.

free parameters (4)
  • z0 = not stated
    Regulator for initial embedding into AdS; the paper claims the lift has a finite limit as z0->0 but does not prove it or specify the value used (Section 4.1).
  • k_lift = 5 (MNIST/VOC), 16 (Shapes), 1 (Ising)
    Number of nearest neighbors used to compute the AdS center of mass; chosen per task in Section B.3, and the Ising result with k_lift=1 is the one reported in the main text.
  • k_con = 16 (MNIST/VOC/Shapes), N (Ising)
    Graph connectivity in message passing; N (fully connected) gives best in-distribution Ising performance, and generalization behavior changes with k_con (Appendix B.2).
  • Delta_eps, Delta_sigma = ~1.0, ~0.125
    Trainable conformal dimensions in the Ising output layer (eq. 17); fitted to ground-truth CFT values (Table 2).
assumptions (5)
  • standard math The global conformal group of R^d is isomorphic to PO(d+1,1).
    Invoked in Section 3.1 and proved in Appendix D.4, following Schottenloher [24].
  • standard math Isom(AdS_{d+1}) is PO(d+1,1).
    Used in Section 3.2 to equate conformal transformations with AdS isometries; cited to McKay [30].
  • domain assumption The 2D Ising CFT correlation functions (eqs. 29-30) give the ground-truth target values.
    Used to generate training data in Section 5.2; standard exact results from the solved Ising CFT.
  • ad hoc to paper The AdS center-of-mass z coordinate has a finite limit as z0->0 and encodes the local scale of the point cloud.
    Stated without proof in Section 4.1 (Algorithm 1); the whole embedding procedure depends on it.
  • domain assumption The lifting of features by z^Delta (eq. 12) correctly implements the transformation of a conformal primary of dimension Delta.
    Motivated by AdS/CFT bulk-to-boundary propagators in Section 4.1; assumed for the Ising output (eq. 17).

how reviews work

0 comments
Cite this review

Pith. "Pith review of AdS-GNN -- a Conformally Equivariant Graph Neural Network." pith.science (2026). https://pith.science/paper/2O3SGPWU

@misc{pith2026250512880,
  author       = {Pith},
  title        = {Pith review of: AdS-GNN -- a Conformally Equivariant Graph Neural Network},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2O3SGPWU}},
  note         = {Machine review of arXiv:2505.12880}
}
read the original abstract

Conformal symmetries, i.e.\ coordinate transformations that preserve angles, play a key role in many fields, including physics, mathematics, computer vision and (geometric) machine learning. Here we build a neural network that is equivariant under general conformal transformations. To achieve this, we lift data from flat Euclidean space to Anti de Sitter (AdS) space. This allows us to exploit a known correspondence between conformal transformations of flat space and isometric transformations on the AdS space. We then build upon the fact that such isometric transformations have been extensively studied on general geometries in the geometric deep learning literature. We employ message-passing layers conditioned on the proper distance, yielding a computationally efficient framework. We validate our model on tasks from computer vision and statistical physics, demonstrating strong performance, improved generalization capacities, and the ability to extract conformal data such as scaling dimensions from the trained network.

Figures

Figures reproduced from arXiv: 2505.12880 by the authors.

Figure 1
Figure 1. AdS-GNN lifts points from Euclidean space [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Example of hyperboloid ∥Y ∥ = −1 embedded in R 2,1 , consti￾tuting two copies of AdS2. To organize the data, we will instead lift it from R d to an auxiliary space with one higher dimension, i.e. Anti de Sitter space AdSd+1. This space can conveniently be understood in terms of a submanifold of R d+1,1 , equipped with its natural metric diag(−1, 1, 1 · · · , 1). Consider the submanifold given by the constraint ∥Y ∥ … view at source ↗
Figure 3
Figure 3. An example of the embedding of a boundary digit into AdS. We will now describe how to use these ideas to formulate a conformally equivariant neural network by extending the data from the boundary into the bulk of AdSd+1. It is possible to do this formally for conformal fields ϕ(x a ) liv￾ing on R d , resulting in a field Φ(x a ; z) living on AdSd+1. However our preliminary experiments show that discretiza￾tion issue… view at source ↗
Figures from the paper (3 more)
Figure 5
Figure 5. Figure 5: Performance on shape segmentation and the Ising task. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Visualization of the output from various models on the Ising task; here all points are fixed [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: Generalization across system size; each square shows the relative L2 loss of a system [PITH_FULL_IMAGE:figures/full_fig_p015_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

56 extracted references · 41 canonical work pages

  1. [1]

    Shape representation in the inferior temporal cortex of monkeys,

    N. K. Logothetis, J. Pauls, and T. Poggio, “Shape representation in the inferior temporal cortex of monkeys,”Current biology5no. 5, (1995) 552–563

    work page 1995

  2. [2]

    Scale and translation-invariance for novel objects in human vision,

    Y . Han, G. Roig, G. Geiger, and T. Poggio, “Scale and translation-invariance for novel objects in human vision,”Scientific reports10no. 1, (2020) 1411

    work page 2020

  3. [3]

    Cardy,Scaling and renormalization in statistical physics, vol

    J. Cardy,Scaling and renormalization in statistical physics, vol. 5. Cambridge university press, 1996

    1996

  4. [4]

    Di Francesco, P

    P. Di Francesco, P. Mathieu, and D. Senechal,Conformal Field Theory. Graduate Texts in Contemporary Physics. Springer-Verlag, New York, 1997

    1997

  5. [5]

    Scale and Conformal Invariance in Quantum Field Theory,

    J. Polchinski, “Scale and Conformal Invariance in Quantum Field Theory,”Nucl. Phys. B303 (1988) 226–236

    work page 1988

  6. [6]

    Scale invariance vs conformal invariance,

    Y . Nakayama, “Scale invariance vs conformal invariance,”Phys. Rept.569(2015) 1–93, arXiv:1302.0884 [hep-th]

    arXiv 2015

  7. [7]

    2d-shape analysis using conformal mapping,

    E. Sharon and D. Mumford, “2d-shape analysis using conformal mapping,”International Journal of Computer Vision70no. 1, (2006) 55–75

    work page 2006

  8. [8]

    Computational conformal geometric methods for vision,

    N. Lei, F. Luo, S.-T. Yau, and X. Gu, “Computational conformal geometric methods for vision,” inHandbook of Mathematical Models and Algorithms in Computer Vision and Imaging: Mathematical Imaging and Vision, pp. 1739–1790. Springer, 2023

    work page 2023

Show all 56 references
  1. [9]

    Group equivariant convolutional networks,

    T. Cohen and M. Welling, “Group equivariant convolutional networks,” inInternational conference on machine learning, pp. 2990–2999, PMLR. 2016

    2016

  2. [10]

    Geometric deep learning: Grids, groups, graphs, geodesics, and gauges,

    M. M. Bronstein, J. Bruna, T. Cohen, and P. Veliˇckovi´c, “Geometric deep learning: Grids, groups, graphs, geodesics, and gauges,”arXiv preprint arXiv:2104.13478(2021)

    2021 arXiv

  3. [11]

    Weiler, P

    M. Weiler, P. Forré, E. Verlinde, and M. Welling,Equivariant and Coordinate Independent Convolutional Networks. 2023.https://maurice-weiler.gitlab.io/cnn_book/ EquivariantAndCoordinateIndependentCNNs.pdf

    2023

  4. [12]

    Implicit convolutional kernels for steerable CNNs,

    M. Zhdanov, N. Hoffmann, and G. Cesa, “Implicit convolutional kernels for steerable CNNs,” inThirty-seventh Conference on Neural Information Processing Systems (NeurIPS). 2023. https://openreview.net/forum?id=2YtdxqvdjX

    2023

  5. [13]

    Clifford-steerable convolutional neural networks,

    M. Zhdanov, D. Ruhe, M. Weiler, A. Lucic, J. Brandstetter, and P. Forré, “Clifford-steerable convolutional neural networks,” inForty-first International Conference on Machine Learning. 2024.https://openreview.net/forum?id=XTglHJjzQI

    2024

  6. [14]

    On the utility of equivariance and symmetry breaking in deep learning architectures on point clouds,

    S. Vadgama, M. M. Islam, D. Buracus, C. Shewmake, and E. Bekkers, “On the utility of equivariance and symmetry breaking in deep learning architectures on point clouds,”arXiv preprint arXiv:2501.01999(2025)

    2025 arXiv

  7. [15]

    Does equivariance matter at scale?,

    J. Brehmer, S. Behrends, P. de Haan, and T. Cohen, “Does equivariance matter at scale?,”arXiv preprint arXiv:2410.23179(2024)

    2024 arXiv

  8. [16]

    Lie group decompositions for equivariant neural networks,

    M. Mironenco and P. Forré, “Lie group decompositions for equivariant neural networks,” inThe Twelfth International Conference on Learning Representations. 2024. https://openreview.net/forum?id=p34fRKp8qA

    2024

  9. [17]

    Lie neurons: Adjoint-equivariant neural networks for semisimple lie algebras,

    T.-Y . Lin, M. Zhu, and M. Ghaffari, “Lie neurons: Adjoint-equivariant neural networks for semisimple lie algebras,” inInternational Conference on Machine Learning, pp. 30529–30545, PMLR. 2024

    2024

  10. [18]

    B-spline cnns on lie groups,

    E. J. Bekkers, “B-spline cnns on lie groups,” inInternational Conference on Learning Representations. 2020.https://openreview.net/forum?id=H1gBhkBFDH

    2020

  11. [19]

    Scale-equivariant steerable networks,

    I. Sosnovik, M. Szmaja, and A. Smeulders, “Scale-equivariant steerable networks,”arXiv preprint arXiv:1910.11093(2019) . 10

    2019 arXiv

  12. [20]

    Truly scale-equivariant deep nets with fourier layers,

    M. A. Rahman and R. A. Yeh, “Truly scale-equivariant deep nets with fourier layers,”Advances in Neural Information Processing Systems36(2023) 6092–6104

    2023

  13. [21]

    Scale-space filtering,

    A. P. Witkin, “Scale-space filtering,” inReadings in computer vision, pp. 329–332. Elsevier, 1987

    1987

  14. [22]

    Deep scale-spaces: Equivariance over scale,

    D. Worrall and M. Welling, “Deep scale-spaces: Equivariance over scale,”Advances in Neural Information Processing Systems32(2019)

    2019

  15. [23]

    Conformal Fields from Neural Networks,

    J. Halverson, J. Naskar, and J. Tian, “Conformal Fields from Neural Networks,” arXiv:2409.12222 [hep-th]

  16. [24]

    Schottenloher,A Mathematical Introduction to Conformal Field Theory

    M. Schottenloher,A Mathematical Introduction to Conformal Field Theory. Lecture Notes in Physics. Springer: Berlin, Heidelberg, 2nd ed., 2008

    2008

  17. [25]

    The Conformal Bootstrap,

    D. Simmons-Duffin, “The Conformal Bootstrap,” inTheoretical Advanced Study Institute in Elementary Particle Physics: New Frontiers in Fields and Strings, pp. 1–74. 2017. arXiv:1602.07982 [hep-th]

    2017 arXiv

  18. [26]

    What Maxwell Theory in D<>4 teaches us about scale and conformal invariance,

    S. El-Showk, Y . Nakayama, and S. Rychkov, “What Maxwell Theory in D<>4 teaches us about scale and conformal invariance,”Nucl. Phys. B848(2011) 578–593, arXiv:1101.5385 [hep-th]

    2011 arXiv

  19. [27]

    The renormalization group and critical phenomena,

    K. G. Wilson, “The renormalization group and critical phenomena,”Reviews of Modern Physics 55no. 3, (1983) 583

    1983

  20. [28]

    Kardar,Statistical physics of fields

    M. Kardar,Statistical physics of fields. Cambridge University Press, 2007

    2007

  21. [29]

    Zinn-Justin,Quantum field theory and critical phenomena, vol

    J. Zinn-Justin,Quantum field theory and critical phenomena, vol. 171. Oxford university press, 2021

    2021

  22. [30]

    An introduction to Cartan geometries,

    B. McKay, “An introduction to Cartan geometries,”arXiv preprint arXiv:2302.14457(2023)

    2023 arXiv

  23. [31]

    The Large N limit of superconformal field theories and supergravity,

    J. M. Maldacena, “The Large N limit of superconformal field theories and supergravity,”Int. J. Theor. Phys.38(1999) 1113–1133,arXiv:hep-th/9711200 [hep-th]. [Adv. Theor. Math. Phys.2,231(1998)]

    1999 arXiv

  24. [32]

    Gauge theory correlators from noncritical string theory,

    S. S. Gubser, I. R. Klebanov, and A. M. Polyakov, “Gauge theory correlators from noncritical string theory,”Phys. Lett. B428(1998) 105–114,arXiv:hep-th/9802109

    1998 arXiv

  25. [33]

    Anti de Sitter space and holography,

    E. Witten, “Anti de Sitter space and holography,”Adv. Theor. Math. Phys.2(1998) 253–291, arXiv:hep-th/9802150

    1998 arXiv

  26. [34]

    A concept of the mass center of a system of material points in the constant curvature spaces,

    G. Galperin, “A concept of the mass center of a system of material points in the constant curvature spaces,”Communications in Mathematical Physics154(1993) 63–84

    1993

  27. [35]

    The Holographic bound in anti-de Sitter space,

    L. Susskind and E. Witten, “The Holographic bound in anti-de Sitter space,” arXiv:hep-th/9805114

    arXiv

  28. [36]

    E(n) equivariant graph neural networks,

    V . G. Satorras, E. Hoogeboom, and M. Welling, “E(n) equivariant graph neural networks,” in ICML, M. Meila and T. Zhang, eds. 2021

    2021

  29. [37]

    Clifford Group Equivariant Simplicial Message Passing Networks,

    C. Liu, D. Ruhe, F. Eijkelboom, and P. Forré, “Clifford Group Equivariant Simplicial Message Passing Networks,”International Conference on Learning Representations (ICLR)(2024)

    2024

  30. [38]

    Fast, expressive se(n) equivariant networks through weight-sharing in position-orientation space,

    E. J. Bekkers, S. P. Vadgama, R. Hesselink, P. A. van der Linden, and D. W. Romero, “Fast, expressive se(n) equivariant networks through weight-sharing in position-orientation space,” in ICLR. 2024

    2024

  31. [39]

    Geometric deep learning on graphs and manifolds using mixture model cnns,

    F. Monti, D. Boscaini, J. Masci, E. Rodolà, J. Svoboda, and M. M. Bronstein, “Geometric deep learning on graphs and manifolds using mixture model cnns,” inCVPR. 2017

    2017

  32. [40]

    Long range graph benchmark,

    V . P. Dwivedi, L. Rampásek, M. Galkin, A. Parviz, G. Wolf, A. T. Luu, and D. Beaini, “Long range graph benchmark,” inNeurIPS. 2022. 11

    2022

  33. [41]

    Decoupled weight decay regularization,

    I. Loshchilov and F. Hutter, “Decoupled weight decay regularization,” inICLR. 2019

    2019

  34. [42]

    Splinecnn: Fast geometric deep learning with continuous b-spline kernels,

    M. Fey, J. E. Lenssen, F. Weichert, and H. Müller, “Splinecnn: Fast geometric deep learning with continuous b-spline kernels,” inCVPR. 2018

    2018

  35. [43]

    Graph convolutional gaussian processes,

    I. Walker and B. Glocker, “Graph convolutional gaussian processes,” inICML. 2019

    2019

  36. [44]

    Graph attention networks,

    P. Velickovic, G. Cucurull, A. Casanova, A. Romero, P. Liò, and Y . Bengio, “Graph attention networks,” inICLR. 2018

    2018

  37. [45]

    Probabilistic numeric convolutional neural networks,

    M. A. Finzi, R. Bondesan, and M. Welling, “Probabilistic numeric convolutional neural networks,” inICLR. 2021

    2021

  38. [46]

    Conformal Linear Transformations,

    A. R. Amir-Moéz, “Conformal Linear Transformations,”Mathematics Magazine40no. 5, (1967) 268–270. 12 AAdS d+1 Details We provide a few further details about aspects ofAdSd+1: A.1 Action of group Recall that AdSd+1 can be understood as the hyperboloid defined in equation 4. The ...

    1967

  39. [47]

    A similar relation holds for the energy operator with∆ϵ = 1

    The power-law functional form is completely fixed by scale-invariance, and the only input from the theory here is the value of the conformal dimension ∆σ. A similar relation holds for the energy operator with∆ϵ = 1. In this work we build a neural network to predict theN-point ...

  40. [48]

    as the groupConfDiff(Rp,q) of all conformal diffeomorphisms ofRp,q, which would include all affine conformal transformations, but exclude the inversion and the special conformal transformations

  41. [49]

    as the group ConfDiff(Mp,q) of all conformal diffeomorphisms of a conformal compactifi- cation Mp,q of Rp,q, which would include all affine and special conformal transformations and the inversion. However, one can show that for p+q= 2 some of the properties of the conformal co...

  42. [50]

    as the connected component of the identity of one of the above groups, in order to stick to the main conformal diffeomorphisms that have descriptions with help of a Lie algebra and Lie exponential map, etc

  43. [51]

    as the subgroup of (partially defined) conformal diffeomorphisms of Rp,q that is generated by all affine and special conformal transformations (excluding the inversion) ofRp,q

  44. [52]

    For more details see further below

    as the subgroup of conformal diffeomorphisms of Mp,q that is generated by all affine and special conformal transformations (excluding the inversion) ofRp,q; For this draft we will settle with one of the last definitions, which are mostly equivalent, and call it the (restricted...

  45. [53]

    We introduce theaffine zero quadricas the following subspaceY 0 ofRp+1,q+1 via: Y0 := y∈R p+1,q+1\{0} ηp+1,q+1(y,y) = 0 ,(134) and endow the tangent spaces fory∈Y 0: TyY0 = v∈R p+1,q+1 ηp+1,q+1(y,v) = 0 ⊆R p+1,q+1,(135) with the pull-back metricη Y0 ofη p+1,q+1: ηY0 y (v1,v 2)...

  46. [54]

    We also introduce thedouble sphere: Sp,q :=S p×Sq :=   y∈R p+1,q+1 pX i=0 |yi|2 = d+1X j=p+1 |yj|2 = 1   ⊆Y 0,(137) and endow the tangent spaces fory∈S p,q: TySp,q =   v∈R p+1,q+1 pX i=0 yi·vi = d+1X j=p+1 yj·vj = 0   ⊆R p+1,q+1.(138) with the pull-back metricη Sp,...

  47. [55]

    (178) i.e

    Letp+q≥3 , then for every conformal mapφ:U→R p,q, defined on any connected open subsetU⊆R p,q, there exists an (up to sign) unique matrix±Λ∈O(p+ 1,q+ 1) such that the following diagram commutes: U [ι] φ // Rp,q [ι] Mp,q [Λ] // Mp,q. (178) i.e. for allx∈Uwe have: [ι◦φ(x)] = [Λι...

  48. [56]

    1−η(ς(x),ς(x)) 2ς(x) 1 +η(ς(x),ς(x)) # (183) = 1 2   1− 1 η(x,x) 2x η(x,x) 1 + 1 η(x,x)   (184) = 1 2 1 η(x,x)

    Let (p,q) = (2,0) , then for everyinjectiveconformal map φ:U→R 2,0, defined either on U=R 2,0 or on anypunctured plane U=R 2,0\{˜x}, there exists an (up to sign) unique matrix±Λ∈O(3,1)such that the corresponding diagram from 178 commutes. 13 Remark D.4.4.For (p,q) = (1,1) the ...

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.