Pith. sign in

REVIEW 3 major objections 5 minor 15 references

Thickness-aware E(3)-Equivariant 3D Mesh Neural Networks

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read By adding one gated edge per node to its pair on the opposite face, T-EMNN predicts node-level 3D deformation on real industrial parts more accurately than surface-only mesh networks while keeping the low cost of a surface mesh.

desk verdict Useful thickness-aware message passing undermined by a flawed E(3)-invariance proof. read the letter →

arxiv 2505.21572 v1 pith:UZIS4DVA submitted 2025-05-27 cs.CV cs.AIcs.LG

classification cs.CVcs.AIcs.LG
keywords meshneuralnetworksthickness-awaremessagepassingE(3)-equivariancenode-leveldeformationpredictiondata-drivencoordinatesystemPCAprincipalaxesinjectionmoldingdatasetsurrogatemodeling
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

Surface meshes capture the outer shell of a 3D object but not what happens across its thickness, and the paper argues this is a real blind spot: the two faces of a thick part deform in near lockstep, so a network that never connects opposing surfaces is missing the object's most useful signal. T-EMNN fixes this by adding a single thickness edge per node, linking it to the nearest node on the opposite face along the inward normal, and gating that edge with a sigmoid whose learnable threshold separates genuine thickness from in-plane width. The paper also introduces a data-driven coordinate frame centered on the mesh and aligned by PCA principal axes, with a sign rule that makes node coordinates E(3)-invariant; predictions are mapped back to the original frame so the whole model is E(3)-equivariant. On a real injection-molding dataset with roughly 54,000 nodes per mesh, T-EMNN reports the lowest RMSE and highest R2 among MGN, EGNN, and EMNN baselines, in both the original and randomly rotated test settings, and it does so with GPU memory close to the cheaper MGN baseline.

What carries the argument

The load-bearing devices are the thickness edge and the data-driven coordinate system. A thickness edge connects a node vi to T(vi), the nearest node found by projecting a ray along the inward normal onto the opposing surface; its feature is the pair [t(vi), n_i·n_T], where t(vi) is the Euclidean distance between the paired nodes and the dot product measures normal alignment. Message passing along the edge is multiplied by the activation I_i = 1/(1 + exp(α(t(vi) − τ))) with α = 3, so edges below the learnable threshold τ pass nearly full weight and edges above it are masked, and this is what separates thickness from width. The coordinate system works by subtracting the center of mass, computing the PCA basis of the centered point cloud as the rotation matrix, flipping each axis to align with v = x_cm − x_bbox, and transforming every coordinate into that frame; the same stored rotation and center transform predictions back, which gives E(3)-equivariance.

What would settle it

Take a mesh with an n-fold rotational symmetry axis, such as a long cylinder or a regular prism, apply a rotation by 360/n degrees, and recompute x_inv via the paper's four steps; if the covariance matrix has repeated eigenvalues, the principal axes are not unique, and the resulting x_inv for corresponding nodes will differ, showing the claimed E(3)-invariance fails. An even simpler check uses a cube: for a uniformly sampled cube, the center of mass equals the bounding-box center, so the reference vector v in Eq. 5 is zero, the sign rule is undefined, and rotating the cube by any angle that permutes the principal axes will change the computed coordinates.

Watch

Extended reading notes

Core claim

The central claim is that thickness is not a side detail of mesh geometry but a first-class interaction: nodes on opposite faces of a solid are so tightly coupled that their deformations are far more correlated than any radius-based neighborhood, and modeling that coupling directly improves prediction. The paper demonstrates the correlation empirically and then shows that a single learned edge per node—connecting each node to the closest node on the opposing surface—is enough to exploit it, provided the edge is gated by a learnable thickness threshold so that pairs measuring 'width' rather than 'thickness' are filtered out. Equally central is the claim that raw spatial coordinates can be fed into an E(3)-equivariant network without expensive spherical harmonics: centering the mesh, taking PCA principal axes as a rotation, and fixing axis signs with a bounding-box reference yields a coordinate frame that is invariant to rotation, translation, and reflection, and the stored rotation and center make the decoder's output equivariant. The paper backs these claims with a real-world injection-molding dataset and reports that the full model outperforms all baselines on node-level 3D deformation while adding only a small computational overhead.

Load-bearing premise

The data-driven coordinate system is only E(3)-invariant if the shape's principal axes are uniquely determined and the sign rule resolves their direction; for rotationally symmetric shapes the PCA eigenspace is degenerate and the sign can stay ambiguous, in which case the rotation of the object can change the computed invariant coordinates.

Editorial extensions

If this is right

  • Adding thickness edges improves RMSE and R2 for node-level 3D deformation prediction on a real industrial dataset, and the gain holds when test inputs are randomly rotated, because the data-driven coordinate frame keeps spatial features aligned.
  • The learned threshold τ converges to a stable value (~5.68) across seeds and matches the best fixed-threshold sweep; removing thickness edges or setting the threshold too high (admitting width pairs) degrades performance.
  • Plugging the same thickness edges and thickness processor into MGN, EGNN, and EMNN improves all three baselines, so the mechanism is a plug-in, not a redesign.
  • On the Deforming Plate dynamics benchmark, thickness edges also improve next-timestep deformation prediction, showing the idea carries over from static analysis to time-dependent simulation.
  • The approach keeps the computational profile of a surface mesh: T-EMNN trains at about 20 iterations per second and uses roughly 3.7 GB GPU memory, far below EMNN's 7.3 GB.

Reading between the lines

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

  • A testable extension is to replace the single scalar threshold with per-region or per-node thresholds, since real parts can have locally varying thickness; the paper's global τ may under-use thin regions or over-include thick ones.
  • The thickness-edge idea is really a cheap way to add one specific long-range interaction, so it could be adapted to other paired physical couplings (contact pairs, symmetry planes, boundary layers) where surface-only message passing would need many hops to connect the two sides.
  • The PCA coordinate frame's equivariance is hostage to the uniqueness of principal axes; for shapes with rotational symmetry the covariance eigenspace is degenerate, so a robust variant might predict jointly from all equivalent principal-axis choices or average invariant features, though that would change the architecture rather than the proof.
  • Because the paper computes thickness pairs once per mesh and keeps them fixed over time in the dynamic experiment, a natural stress test is whether recomputing pairs as the mesh deforms changes predictions; if it does, the frozen-pair assumption is doing some of the work.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. This paper proposes T-EMNN, a mesh-based graph neural network for predicting node-level 3D deformation of objects with thickness. Two main components are introduced: (i) thickness edges that connect each surface node to a paired node on the opposing surface, with a learnable threshold and sigmoid activation to retain only 'true thickness' pairs; and (ii) a data-driven coordinate system built from the PCA of the shape's point cloud, intended to give E(3)-invariant spatial coordinates. The method is evaluated on a real-world injection-molding dataset and the Deforming Plate benchmark, reporting improved RMSE/MAE/R² over MGN, EGNN, and EMNN baselines, including under randomly rotated test inputs.

Significance. If the E(3)-invariance guarantee held, the data-driven coordinate system would be an attractive low-cost way to inject global spatial information into equivariant mesh networks, and the thickness-edge message passing is a simple, well-motivated mechanism for capturing through-thickness correlations that surface-only meshes miss. The experimental study is unusually careful: Table 1 separates the effects of spatial information and thickness edges, Fig. 6 validates the learned threshold by scanning fixed thresholds, Fig. 8 shows thickness edges help all baselines, and Table 4 documents computational cost. The main weakness is that the central theoretical claim--E(3)-invariance of the PCA-based coordinates--is not actually proved and, as stated, is false; this must be corrected or the claims substantially weakened before the contribution is complete.

major comments (3)
  1. [Sec. 4.1, Step 3 / Appendix H] The sign-alignment rule in Eq. (5) is not rotation-equivariant, so the claimed E(3)-invariance of x_inv does not hold even for generic non-degenerate shapes. The reference vector v = x_cm - x_bbox depends on the axis-aligned bounding box, and the componentwise min and max operations do not commute with rotations: under a rotation Q, v' != Q v in general. A concrete 2D example is X = {(0,0),(2,0),(0,2),(3,1)}; rotating the cloud by 30 degrees changes the sign of both PCA axes relative to the unrotated aligned frame, giving x_inv' = -x_inv. The proof in Appendix H explicitly states 'we assume that the signs of the basis vectors are already aligned,' which is precisely the step that fails, and Eq. (25) also uses an invalid matrix product (the covariance of the rotated data should be Q C_X Q^T, not C_X). Because x_inv is an input to the spatial encoder (Eq. 10) and decoder (Eq. 19), a sign flip changes the network's inputs and cannot be canceled by the inverse rotation unless the network is sign-equivariant, which is not the case for an MLP. Please replace Eq. (5) with a reference vector that transforms as a vector under E(3) (e.g., based on the farthest point from the center of mass) or substantially weaken the invariance/equivariance claims.
  2. [Sec. 3.3, Eq. (1)] The scalar d ('ray projection distance') is never defined or specified anywhere in the manuscript. Since T(v_i) is the argmin of ||x_j - (x_i - d·n_i)||, the selected thickness pair depends on d: as d approaches 0 the pair is the closest node on the opposite side, while as d grows large it approaches the node with the most negative projection along n_i. No value, tuning procedure, or default is reported in the experiments (Section 5) or the appendix, so the thickness-edge construction is not reproducible and the reported numbers cannot be independently checked. Please report d or justify that the results are insensitive to it over a wide range.
  3. [Sec. 4.2.4, inverse transformation] The final deformation is computed as p_orig_i = R·p_inv_i + x_cm. If p_inv_i is a deformation (displacement) vector predicted in the invariant coordinate system, the translation x_cm should not be added; displacements are vectors and transform under rotation only. Adding x_cm would make the prediction depend on the global position of the shape and is inconsistent with the stated task of predicting node-level deformation. Equation (8) is the correct inverse mapping for absolute coordinates, not for displacements. Please clarify whether the network predicts absolute deformed positions or displacements and correct the formula, or confirm that the reported experiments use R·p_inv_i and the equation is a typo.
minor comments (5)
  1. [Sec. 3.3, Eq. (1)] The strict inequality (x_j - x_i)·n_i < 0 may leave T(v_i) undefined if no node lies strictly on the opposite side; please state the fallback (e.g., ignore the thickness edge for that node).
  2. [Sec. 4.2.3, Eq. (15)] The sigmoid activation I_i is continuous, so edges with t(v_i) > tau are not actually excluded, only down-weighted; the text later says such edges are 'filtered out.' Clarify whether hard masking is applied at test time or whether the description is only approximate.
  3. [Table 4] Table 4 lists 'MLP' twice with identical input column '-'; presumably the second row should specify a coordinate input (x_orig or x_inv).
  4. [Appendix H, Eqs. (24)-(25)] The notation QX is dimensionally inconsistent when X is an N x 3 matrix; the covariance of rotated data should be expressed as Q C_X Q^T (or the matrix dimensions must be stated explicitly). This makes the derivation hard to follow.
  5. [Sec. 5.4.3] The dynamic experiment uses the original coordinate system because the data-driven coordinate system is 'designed for static analysis'; this is unexplained, since the coordinate transformation itself does not depend on time, and it weakens the comparison with the static results.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity found: the coordinate transform and thickness processor are derived from input geometry, and the ablations are sensitivity checks, not construction-level fits.

full rationale

The central derivation chain is self-contained. The data-driven coordinate system (Sec. 4.1) is constructed purely from input vertex positions: centering (Eqs. 3-4), PCA axes, sign alignment via the bounding-box reference vector (Eq. 5), and projection (Eq. 7). The target deformation enters only through the training loss; it is never used to define x_inv, R, or the thickness pairs. The E(3)-invariance proof (App. H) is a direct calculation for that construction; it contains an explicit unresolved case, namely "For the proof, we assume that the signs of the basis vectors are already aligned," which is a correctness/robustness gap for symmetric or adversarially rotated shapes, not a circular reduction. The thickness processor is likewise non-circular: T(v_i) and t(v_i) are geometric functions of the mesh (Eqs. 1-2), the threshold tau is a learned model parameter, and the fixed-threshold scan (Fig. 6) is a sensitivity/ablation check, not a prediction forced by construction. The only self-citation (Yu et al. 2023, which includes co-authors Lee and Park) appears in the related-work list of hierarchical pooling methods and is not load-bearing for any claim. No equation reduces to its own input, and no fitted parameter is renamed as a prediction; the paper's flagged limitation is a correctness concern, not circularity.

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

The central claim rests on two modeling choices (thickness edges and the invariant coordinate frame) plus several unstated assumptions: the PCA axes must be unique and signable, the mesh must be water-tight, and the thickness pairing must be robust to the unspecified d parameter. The learned threshold tau and alpha are ordinary hyperparameters, but d is a hidden free parameter that affects the data construction.

free parameters (3)
  • Thickness threshold tau = 5.68
    Learned via backpropagation; controls which thickness edges are active. Fixed-threshold scan in Fig. 6 confirms this value is near-optimal.
  • alpha (scaling factor) = 3
    Chosen by hyperparameter search over {1,3,5,10}; balances smooth masking and precise thresholding, as discussed in Appendix D.
  • d (ray projection distance) = not reported
    Used in Eq. 1 to define the thickness node pair. The paper never specifies how d is set, and the pairing result depends on it, making the method under-specified.
assumptions (3)
  • domain assumption The point cloud used for PCA has non-degenerate principal components and sign alignment always succeeds.
    The E(3)-invariance proof in Appendix H assumes signs of PCA basis vectors are already aligned and does not handle degenerate eigenspaces that occur for symmetric shapes.
  • domain assumption The mesh is water-tight and has consistently oriented outward normals.
    The thickness pairing rule in Eq. 1 relies on an inward normal direction and a unit normal at each node, which requires a closed oriented surface.
  • domain assumption The nearest qualifying node along the inward normal is the physically meaningful opposing point.
    The definition of T(v_i) in Eq. 1 selects the closest node satisfying the dot-product constraint; this is assumed to represent the true opposite surface point, but the choice depends on the unstated distance d.
invented entities (3)
  • Thickness edge
    purpose: Adds a directed connection from a node to its thickness paired node to propagate information between opposing surfaces.
    The benefit is shown only on the private industrial dataset and the public Deforming Plate ablation; no independent benchmark outside this paper is provided.
  • Thickness node pair
    purpose: Defines the pairing between a surface node and the node on the opposite surface, enabling thickness measurement and edge construction.
    The pairing rule is introduced in this paper and its validity is tied to the same datasets used for evaluation.
  • Data-driven coordinate system
    purpose: Replaces raw coordinates with a PCA-based canonical frame to make the network invariant to E(3) transformations.
    The OOD rotation experiments provide in-paper evidence, but no external validation or formal guarantee for degenerate shapes is given.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Thickness-aware E(3)-Equivariant 3D Mesh Neural Networks." pith.science (2026). https://pith.science/paper/UZIS4DVA

@misc{pith2026250521572,
  author       = {Pith},
  title        = {Pith review of: Thickness-aware E(3)-Equivariant 3D Mesh Neural Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UZIS4DVA}},
  note         = {Machine review of arXiv:2505.21572}
}
read the original abstract

Mesh-based 3D static analysis methods have recently emerged as efficient alternatives to traditional computational numerical solvers, significantly reducing computational costs and runtime for various physics-based analyses. However, these methods primarily focus on surface topology and geometry, often overlooking the inherent thickness of real-world 3D objects, which exhibits high correlations and similar behavior between opposing surfaces. This limitation arises from the disconnected nature of these surfaces and the absence of internal edge connections within the mesh. In this work, we propose a novel framework, the Thickness-aware E(3)-Equivariant 3D Mesh Neural Network (T-EMNN), that effectively integrates the thickness of 3D objects while maintaining the computational efficiency of surface meshes. Additionally, we introduce data-driven coordinates that encode spatial information while preserving E(3)-equivariance or invariance properties, ensuring consistent and robust analysis. Evaluations on a real-world industrial dataset demonstrate the superior performance of T-EMNN in accurately predicting node-level 3D deformations, effectively capturing thickness effects while maintaining computational efficiency.

Figures

Figures reproduced from arXiv: 2505.21572 by the authors.

Figure 1
Figure 1. The left figures show a mesh, with two different target nodes (•), their thickness paired nodes (•), thickness distance (−), and nearby nodes within a radius (•). The right figures compare Pearson correlation and L2 Norm between the target node’s defor￾mation and its thickness paired / nearby nodes within a radius. complex systems. Traditionally, computational numerical solvers like finite element methods (FEM) (Klo… view at source ↗
Figure 2
Figure 2. Overview of T-EMNN. 𝐛! 𝐛" 𝐛# 𝐱$$%& 𝐱'( 𝐛# 𝐯 = 𝐱!" − 𝐱##$% + [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Our proposed data-driven coordinate system. system defined by the shape itself, independent of its orien￾tation or alignment in the original coordinate system. This transformation is achieved through the following steps: Step 1: Adjust Coordinates to Center of Mass. The center of mass xcm of the shape is calculated as: xcm = 1 |V | X vi∈V x orig i , (3) where |V | is the total number of nodes in the mesh. Each node’… view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: The concept of thickness (left) and width (right). Sec 3.3. In brief, thickness is characterized by the spatial separation between opposing surfaces, with thickness paired node T (vi) of node vi defined as the closest node on the op￾posing surface along the inward norm…
Figure 5
Figure 5. Figure 5: Learning curve of the thickness threshold τ during train￾ing across three seeds (left), and the distribution of thickness values t(vi) with the cutoff threshold (red dotted line, t(vi) = τ ) used for message passing in the thickness processor (right). with E(3)-transfo…
Figure 7
Figure 7. Figure 7: Visualization of error magnitude (RMSE). The ground truth shows deformation magnitude (a), while (b–f) illustrate prediction errors. Additional examples are in [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: Comparison of baselines with x inv and their extension with our thickness edges. models that originally lacked it, such as (e) MGN. This ca￾pability transforms previously limited models into robust systems capable of handling transformations effectively, thereby greatl…
Figure 10
Figure 10. Figure 10: Visualization of thickness edges in the Deforming Plate dataset: edges with t(vi) below the learned threshold τ (left), and edges with t(vi) above τ , which are filtered out by the thickness processor (right). We first construct thickness edges based on the mesh at th…
Figure 11
Figure 11. Figure 11: Examples of Dataset Shapes. The experimental conditions consist of eight types for each shape: pack pressure, pack time, projected area, gate size, injection value, volume, melt temperature, and mold temperature. By varying these combinations, a total of 18 experiment…
Figure 12
Figure 12. Figure 12: Analysis of Hyperparameter α. Hyperparameter α is a scaling factor that influences the transition sharpness. Specifically, as α increases, the thickness activation value Ii in Eq. 15 discretely masks the thickness edges whose thickness values are near the threshold. I…
Figure 13
Figure 13. Figure 13: R 2 scores for all test data. In the shape IDs, ‘s’ indicates seen shapes included in the training data, while ‘us’ refers to unseen shapes. The number following ‘s’ or ‘us’ represents the shape type, and ‘c’ denotes the experimental condition, followed by its corresp…
Figure 14
Figure 14. Figure 14: Comparisons between volume mesh and surface mesh. The methods used for comparison are based on the MGN framework with coordinate embeddings from our proposed coordinate system. GPU memory usage represents the average GPU consumption across the test dataset, and infere…
Figure 15
Figure 15. Figure 15: Visualization of error magnitude (RMSE). The ground truth represents the magnitude of deformation, while each method’s figure illustrates the prediction error (RMSE) relative to the ground truth. 15 [PITH_FULL_IMAGE:figures/full_fig_p015_15.png]
Figure 16
Figure 16. Figure 16: Detailed visualization of thickness edges with t(vi) below the learned threshold τ (left) and those above τ , filtered out by the thickness processor (right). 16 [PITH_FULL_IMAGE:figures/full_fig_p016_16.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

15 extracted references · 5 canonical work pages

  1. [1]

    and Yahav, E

    Alon, U. and Yahav, E. On the bottleneck of graph neural networks and its practical implications.arXiv preprint arXiv:2006.05205,

  2. [5]

    and Ji, S

    Gao, H. and Ji, S. Graph u-nets. Ininternational conference on machine learning, pp. 2083–2092. PMLR,

  3. [6]

    Directional message passing for molecular graphs.arXiv preprint arXiv:2003.03123,

    Gasteiger, J., Groß, J., and G ¨unnemann, S. Directional message passing for molecular graphs.arXiv preprint arXiv:2003.03123,

  4. [8]

    Eagle: Large-scale learning of turbulent fluid dynamics with mesh transformers.arXiv preprint arXiv:2302.10803,

    Janny, S., Beneteau, A., Nadri, M., Digne, J., Thome, N., and Wolf, C. Eagle: Large-scale learning of turbulent fluid dynamics with mesh transformers.arXiv preprint arXiv:2302.10803,

  5. [14]

    #MGN + 𝐱!

    Comparisons between volume mesh and surface mesh. The methods used for comparison are based on the MGN framework with coordinate embeddings from our proposed coordinate system. GPU memory usage represents the average GPU consumption across the test dataset, and inference time reflects the total time required to process the test dataset. Mesh representatio...

  6. [1991]

    Constraint-based graph network simulator

    Rubanova, Y ., Sanchez-Gonzalez, A., Pfaff, T., and Battaglia, P. Constraint-based graph network simulator. arXiv preprint arXiv:2112.09161,

  7. [2007]

    Pfaff, T., Fortunato, M., Sanchez-Gonzalez, A., and Battaglia, P. W. Learning mesh-based simulation with graph networks.arXiv preprint arXiv:2010.03409,

  8. [2008]

    Pre- dicting physics in mesh-reduced space with temporal attention.arXiv preprint arXiv:2201.09113,

    Han, X., Gao, H., Pfaff, T., Wang, J.-X., and Liu, L.-P. Pre- dicting physics in mesh-reduced space with temporal attention.arXiv preprint arXiv:2201.09113,

Show all 15 references
  1. [2010]

    Generalized beam theory applied to shear stiff- ness.International Journal of Solids and Structures, 27 (15):1955–1967,

    Renton, J. Generalized beam theory applied to shear stiff- ness.International Journal of Solids and Structures, 27 (15):1955–1967,

  2. [2018]

    Learning flexible body collision dynamics with hier- archical contact mesh transformer.arXiv preprint arXiv:2312.12467,

    Yu, Y .-Y ., Choi, J., Cho, W., Lee, K., Kim, N., Chang, K., Woo, C.-S., Kim, I., Lee, S.-W., Yang, J.-Y ., et al. Learning flexible body collision dynamics with hier- archical contact mesh transformer.arXiv preprint arXiv:2312.12467,

  3. [2019]

    Multiscale meshgraphnets.arXiv preprint arXiv:2210.00612,

    Fortunato, M., Pfaff, T., Wirnsberger, P., Pritzel, A., and Battaglia, P. Multiscale meshgraphnets.arXiv preprint arXiv:2210.00612,

  4. [2020]

    Neural operator: Graph kernel network for partial differential equations

    Anandkumar, A., Azizzadenesheli, K., Bhattacharya, K., Kovachki, N., Li, Z., Liu, B., and Stuart, A. Neural operator: Graph kernel network for partial differential equations. InICLR 2020 Workshop on Integration of Deep Neural Models and Differential Equations,

  5. [2021]

    Suk, J., de Haan, P., Lippe, P., Brune, C., and Wolterink, J. M. Equivariant graph neural networks as surrogate for computational fluid dynamics in 3d artery models. In Fourth workshop on machine learning and the physical sciences (NeurIPS 2021),

  6. [2022]

    J., and Welling, M

    Brandstetter, J., Hesselink, R., van der Pol, E., Bekkers, E. J., and Welling, M. Geometric and physical quantities improve e (3) equivariant message passing.arXiv preprint arXiv:2110.02905,

  7. [2023]

    Deeper insights into graph convolutional networks for semi-supervised learning

    Li, Q., Han, Z., and Wu, X.-M. Deeper insights into graph convolutional networks for semi-supervised learning. In Proceedings of the AAAI conference on artificial intelli- gence, volume 32, 2018a. Li, Y ., Wu, J., Tedrake, R., Tenenbaum, J. B., and Torralba, A. Learning partic...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.