Pith. sign in

REVIEW 3 major objections 5 minor 49 references

Effects of relational graph modularity and depth on the learning performance of neural networks

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read This paper claims that the community structure of a neural network's hidden-layer graph improves learning at five layers—by up to 5.3% top-1 error on CIFAR-10—and that this advantage fully reverses at eight layers.

desk verdict Competent empirical extension of graph2nn with a genuine depth-dependent reversal, but the modularity claim needs parameter-matched controls before it is fully clean. read the letter →

arxiv 2507.10005 v2 pith:6HWZ6ER4 submitted 2025-07-14 cs.LG cond-mat.stat-mechcs.NEphysics.comp-ph

Yash Arya , Sang Hoon Lee This is my paper
classification cs.LGcond-mat.stat-mechcs.NEphysics.comp-ph
keywords relationalgraphscommunitystructurenetworkmodularityneuraldepthCIFAR-10scale-freenetworksover-smoothingimageclassification
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

The authors test whether the wiring pattern of a neural network's hidden layers—not just how many connections exist, but how they are arranged into communities—changes how well the network learns. They do this with a relational-graph representation in which each hidden layer is one round of message exchange along a fixed sparse graph, and they build such graphs as random or scale-free networks split into a tunable number of communities, trained on the CIFAR-10 image classification task. They find that at five layers, every sparse graph tested beats the fully connected baseline, and graphs with densely interconnected communities do best, up to 5.3% lower top-1 error. They also find that increasing the density of connections between communities steadily improves performance, with rank correlation reaching −0.91 at eight communities. The pattern is depth-bounded: at eight layers all tested configurations perform worse than the fully connected baseline, which the authors attribute to over-smoothing, the collapse of node representations toward uniformity after many rounds of message passing.

What carries the argument

The central object is the relational graph: a fixed sparse graph whose nodes pair one input and one output channel of a hidden layer, and whose edges define which pairs of neurons exchange messages. Each hidden layer corresponds to one round of message passing over the same sparsity pattern, with independently learned weights per round. The paper's generative machinery is a simplified community-benchmark construction: the network is partitioned into a variable number of communities, each generated as a random or scale-free subgraph, and connected to one another by a tunable inter-community density parameter µ. This design lets the authors sweep modularity while holding layer width and depth fixed, and then map top-1 error over the (intra-community density, µ, number-of-communities) phase space against a fully connected baseline, and correlate error with µ using a non-parametric rank correlation.

What would settle it

Train the same family of relational graphs with the number of trainable weights matched across configurations—for example, by widening the sparse layers or adding untrained connections so every graph has the same parameter count—and check whether the five-layer community advantage and the eight-layer reversal persist; if the advantage disappears, it was driven by parameter reduction, not community structure.

Watch

Extended reading notes

Core claim

The paper's central claim is that modularity of the relational graph is a first-order, depth-bounded determinant of learning performance. On the CIFAR-10 task with a fixed-width five-layer MLP, a sparse relational graph of either random or scale-free type consistently improves top-1 error relative to the complete-graph baseline, and imposing community structure on that graph improves it further, with the best configuration outperforming the complete graph by 5.3%. The paper identifies inter-community density as the dominant structural lever: across configurations with 2 to 8 communities, higher inter-community density is monotonically associated with lower error, with rank correlation reaching −0.91 at 8 communities. The same experiment at eight layers produces a categorical reversal: every tested configuration, including community-structured ones, falls below the complete-graph baseline, and the variance across topologies grows. The authors explain the reversal as over-smoothing, where repeated rounds of message passing drive node representations toward a uniform, information-poor state, and they conclude that five layers is a principled optimum for this architecture and dataset.

Load-bearing premise

The load-bearing assumption is that the performance differences between graph configurations reflect the topology itself, even though the compared networks contain different numbers of trainable weights—the sparse community graphs have far fewer parameters, so part of their advantage could be an ordinary regularization effect.

Editorial extensions

If this is right

  • If modular sparse wiring helps at five layers, network architecture search should treat community structure as a tunable design axis on top of width, depth, and edge count.
  • If the eight-layer reversal is real, depth and topology cannot be optimized independently: a graph that is beneficial at moderate depth becomes harmful at greater depth, so depth sweeps are required before any wiring claim is adopted.
  • If inter-community density is the dominant lever, then for a fixed edge budget, wiring that distributes edges across communities—rather than only within them—should be prioritized for fixed-width classifiers.
  • A direct corollary is that full connectivity is not the best use of parameters at this scale, so sparse structured layers can be adopted as a default in small fixed-width MLPs.

Reading between the lines

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

  • Beyond the paper: if parameter counts were matched across configurations, the community advantage at five layers might vanish entirely, because edge count alone—not modularity—could explain the gains; a matched-parameter experiment would settle this.
  • Beyond the paper: the strong monotonic link between inter-community density and lower error implies a practical, untested recipe—raising inter-community integration at moderate depth is a cheap performance knob even when the overall edge count is fixed.
  • Beyond the paper: the roundworm brain network outperforming its random subsample suggests that global community organization, not just local wiring, carries the benefit; a further untested extension would ablate communities by targeted edge rewiring to localize where the performance loss concentrates.
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. The manuscript studies how the topology of a fixed-width MLP's hidden layers, represented as a relational graph, affects CIFAR-10 classification performance. The authors generate Erdős–Rényi and static scale-free networks with and without community structure (using a simplified LFR benchmark), translate them into 5-layer and 8-layer message-passing MLPs, and measure top-1 error against a fully connected baseline. They report that community-structured sparse graphs improve over the complete baseline by up to 5.3% at five layers, that inter-community density mu correlates strongly with lower error, that the advantage reverses at eight layers, and that a C. elegans neural network outperforms frontal and ER baselines. They interpret the results as evidence for a modular inductive bias and structural regularization.

Significance. If the conclusions were established, the paper would provide a valuable systematic map of how mesoscale community structure affects learning in a controlled relational-graph setting, with potential design guidance for architecture search and biologically inspired networks. The study is empirically extensive for a 128-node setting, includes multiple random seeds with a cross-check at ten seeds, and ships code and an interactive visualization. The main weaknesses are that the parameter-count confound prevents the 5-layer advantage from being attributed to topology, and the 8-layer reversal is demonstrated only for a restricted subset of the parameter space. These issues are fixable with additional matched controls, so the work is a worthwhile candidate for major revision.

major comments (3)
  1. [Sec. II, Eq. (1); Secs. IV.A–IV.B] The experimental design does not match the number of trainable parameters across configurations. Because each message-passing round r has an independent weight matrix W^(r) whose sparsity pattern is fixed by the graph adjacency, the hidden-layer parameter count is proportional to the number of edges |E| in the relational graph. The complete-graph baseline has N^2 weights per round, while every sparse configuration in Figs. 2–4 has fewer. The 5-layer claim that community-structured graphs outperform the complete baseline by up to 5.3% top-1 error (Sec. IV.B) is therefore indistinguishable from a capacity-regularization effect, and the manuscript itself invokes capacity reduction in Sec. IV.A ('reducing model capacity relative to a fixed training set size improves generalization'). The later modularity claims (Secs. IV.B and IV.E) would require either edge-matched non-community baselines or an analysis that holds the parameter count fixed while varying community organization; otherwise the monotonic dependence on inter-community density mu in Fig. 7 could be an edge-count effect rather than a community-topology effect.
  2. [Sec. IV.D, Fig. 6] The categorical depth-reversal claim is not supported by the presented evidence. The 8-layer results in Fig. 6(a) are for static scale-free networks with communities at c=3 and a single fixed inter-community density mu=0.667; Fig. 6(b) is also restricted to c=3. No 8-layer results are shown for ER networks, for other community counts, or for the no-community baselines. The abstract and Sec. V state that extending the architecture to eight layers 'reverses the effect entirely,' but such a categorical conclusion requires 8-layer sweeps over the same parameter space used for the 5-layer phase diagrams. The over-smoothing interpretation is plausible, but the paper does not measure representation collapse or compare against complete-graph over-smoothing; it only cites Refs. [43,44].
  3. [Sec. IV.E, Fig. 7] The Spearman correlations in Fig. 7 are computed after aggregating across all intra-community parameters p (or gamma), and the p range reported in the Fig. 7 caption (p∈[0.167,1.0]) differs from the range stated in the text for Fig. 3 (p∈[0.1,1.0]). Because mu also changes the expected edge count, the correlation does not isolate community topology from network density. Moreover, a Spearman rho of -0.91 does not mean that 'nearly 91% of the variance' is explained (Sec. IV.E); rho is a rank correlation, not an R^2. The authors should report partial correlations controlling for edge count or hold p/gamma fixed, and should rephrase the variance claim.
minor comments (5)
  1. [Fig. 4 and Fig. 7 captions] The captions of Fig. 4 and Fig. 7 are essentially identical, although the figures are referred to in different sections; one of the captions likely describes the wrong figure.
  2. [Sec. II] There is a typo in the phrase 'thesmall world” property' — the opening quotation mark and a space are missing.
  3. [Sec. IV.B] The sentence 'the implementation of any relational graph generally enhances the MLP architecture’s predictive performance by 2.8%' is ambiguous: it should specify whether 'any' means every tested graph, and whether 2.8% is a mean improvement over the baseline.
  4. [Footnote 31, Sec. III.C] The footnote states that only the largest connected component of each community is retained, so the actual node count and edge count vary across configurations; this should be reported explicitly in the methods because it affects both the parameter-count analysis and the comparison across community numbers.
  5. [Sec. IV.C, Fig. 5] The C. elegans comparison in Fig. 5 reports small differences (about 0.3–0.6 percentage points) with error bars; the text should state which pairwise differences are statistically significant and should clarify that the 'Random' sample is not matched to the frontal network in degree sequence or edge density.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper is an empirical benchmark study; the reported structural effects are measured outcomes, not quantities derived from fitted inputs.

full rationale

The paper's central claims are observational: it sweeps graph-generation parameters (p, gamma, mu, c), trains fixed-width MLPs, and reports top-1 error rates. There is no fitted parameter that is later relabeled as a prediction, and no equation is defined in terms of its own output. The relational-graph construction (Eq. 1) is adopted from an external prior work (Ref. [13]) rather than introduced here, and the depth-reversal result is compared against the same external paper's documented failure cases plus independent over-smoothing references (Refs. [43,44]); none of these are self-citations. The only self-citation (Ref. [49], a prior paper by author S.H. Lee) appears in the acknowledgments as an 'initial assessment' of graph2nn and is not load-bearing for any argument. The chief weakness--trainable parameter counts are not matched across sparse and complete configurations, so the reported advantage of sparse/community graphs may partly reflect capacity regularization rather than topology--is a genuine confound affecting causal attribution, but it is not circularity under the definitions used here. Similarly, comparing 8-layer sparse configurations against a 5-layer complete-graph baseline mixes depth and topology, but again this is an experimental validity concern rather than a derivation that reduces to its inputs. The reported Spearman correlations (e.g., rho approx -0.91 for mu at c=8) are summary statistics of the experiments, not predictions forced by construction.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical or mathematical entities. It relies on standard network models (ER, static scale-free, simplified LFR) and a graph-to-network mapping from prior work. The main assumptions are the faithfulness of the relational graph representation, the applicability of bias-variance and over-smoothing theory, and the handling of disconnected components.

free parameters (3)
  • intra-community connection probability p (ER) / degree exponent gamma (scale-free) = varied over p in [0.1,1.0] and gamma in [2,6]
    These are the network-generation parameters scanned to produce the phase diagrams; they are chosen by the authors, not fitted to the data, and their effect on performance is the object of study.
  • inter-community density mu = varied over [0,1]
    The mixing parameter that controls the fraction of edges connecting different communities; it is swept and reported as the main performance driver.
  • average degree m (scale-free networks) = m=3 for community experiments
    Selected from the 5-layer sweet spot in Fig. 2; treated as a fixed design choice for the community and depth experiments.
assumptions (4)
  • domain assumption Relational graph message passing of Eq. (1) faithfully represents the learning dynamics of an MLP layer.
    Invoked in Sec. II; the entire method rests on this equivalence between graph topology and layer connectivity, taken from Ref. 13.
  • domain assumption Statistical learning theory (bias-variance trade-off) explains the superior performance of sparser networks.
    Invoked in Sec. IV.A; the paper explains the sparse advantage via reduced model capacity, but does not verify that variance reduction dominates bias increase in this setting.
  • domain assumption Over-smoothing theory for GNNs applies to these fixed-width MLP relational graphs at 8 layers.
    Invoked in Sec. IV.D-E; the reversal at 8 layers is attributed to over-smoothing based on Refs. 43-44, but no representation-collapse measurement is shown.
  • domain assumption The largest connected component of each generated community is a sufficient proxy for the intended community structure.
    Stated in footnote 31; nodes outside the largest component are discarded, which may distort the intended community sizes and degree distributions.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Effects of relational graph modularity and depth on the learning performance of neural networks." pith.science (2026). https://pith.science/paper/6HWZ6ER4

@misc{pith2026250710005,
  author       = {Pith},
  title        = {Pith review of: Effects of relational graph modularity and depth on the learning performance of neural networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6HWZ6ER4}},
  note         = {Machine review of arXiv:2507.10005}
}
read the original abstract

In recent years, graph-based machine learning techniques, such as reinforcement learning and graph neural networks, have garnered significant attention. While some recent studies have started to explore the relationship between the graph structure of neural networks and their predictive performance, they often limit themselves to a narrow range of model networks, particularly lacking mesoscale structures such as communities. Our work advances this area by conducting a more comprehensive investigation, incorporating realistic network structures characterized by heterogeneous degree distributions and community structures, which are typical characteristics of many real networks. These community structures offer a nuanced perspective on network architecture. Our analysis employs model networks such as random and scale-free networks, alongside a comparison with a biological neural network and its subsets for more detailed analysis. We examine the impact of these structural attributes on the performance of image classification tasks. Our findings reveal that structural properties do affect performance to some extent. Specifically, networks featuring coherent, densely interconnected communities demonstrate enhanced learning capabilities. Crucially, we find that this advantage is depth-dependent: extending the architecture to eight layers reverses the effect entirely. This comparison with the biological neural network emphasizes the relevance of our findings to real-world structures, suggesting an intriguing connection worth further exploration. This study contributes meaningfully to network science and machine learning, providing insights that could inspire the design of more biologically informed neural networks.

Figures

Figures reproduced from arXiv: 2507.10005 by the authors.

Figure 1
Figure 1. FIG. 1. Schematic illustration of the translation between multi-layer perceptron (MLP) architectures and relational graphs. (a) The upper [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2. Performance summary of static scale-free network configu [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3. Performance of Erd [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: FIG. 4. Correlation between inter-community density [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: FIG. 5. Comparison of network performance of the [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: FIG. 6. Comparison of 5-layer and 8-layer fixed-width MLP per [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 7
Figure 7. Figure 7: FIG. 7. Correlation between inter-community density [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

49 extracted references · 46 canonical work pages

  1. [13]

    J. You, J. Leskovec, K. He, and S. Xie, Graph structure of neural networks, inProceedings of the 37th International Conference on Machine Learning, ICML’20 (JMLR.org, 2020)

    work page 2020

  2. [1]

    LeCun, Y

    Y . LeCun, Y . Bengio, and G. Hinton, Deep learning, Nature 521, 436 (2015)

    work page 2015

  3. [2]

    J. You, Z. Ying, and J. Leskovec, Design space for graph neural networks, Advances in Neural Information Processing Systems 33, 17009 (2020)

    work page 2020

  4. [3]

    Zoph and Q

    B. Zoph and Q. V . Le, Neural architecture search with rein- forcement learning, inInternational Conference on Learning Representations (ICLR)(2017)

    work page 2017

  5. [4]

    Dressler and O

    F. Dressler and O. B. Akan, A survey on bio-inspired network- ing, Computer networks54, 881 (2010)

    work page 2010

  6. [5]

    Barab ´asi and R

    A.-L. Barab ´asi and R. Albert, Emergence of scaling in random networks, Science286, 509 (1999)

    work page 1999

  7. [6]

    D. J. Watts and S. H. Strogatz, Collective dynamics of ‘small- world’ networks, Nature393, 440 (1998)

    work page 1998

  8. [7]

    M. A. Porter, J.-P. Onnela, and P. J. Mucha, Communities in networks, Not. Am. Math. Soc.56, 1082 (2009)

    work page 2009

Show all 49 references
  1. [8]

    Fortunato, Community detection in graphs, Phys

    S. Fortunato, Community detection in graphs, Phys. Rep.486, 75 (2010)

    2010

  2. [9]

    Fortunato and M

    S. Fortunato and M. E. J. Newman, 20 years of network com- munity detection, Nature Physics18, 848 (2022)

    2022

  3. [10]

    Bullmore and O

    E. Bullmore and O. Sporns, Complex brain networks: graph theoretical analysis of structural and functional systems, Nature reviews neuroscience10, 186 (2009)

    2009

  4. [11]

    Meunier, R

    D. Meunier, R. Lambiotte, and E. T. Bullmore, Modular and hi- erarchically modular organization of brain networks, Frontiers in neuroscience4, 200 (2010)

    2010

  5. [12]

    Clune, J.-B

    J. Clune, J.-B. Mouret, and H. Lipson, The evolutionary origins of modularity, Proceedings of the Royal Society b: Biological sciences280, 20122863 (2013)

    2013

  6. [14]

    Rosenblatt, The perceptron: A probabilistic model for in- formation storage and organization in the brain, Psychological Review65, 386 (1958)

    F. Rosenblatt, The perceptron: A probabilistic model for in- formation storage and organization in the brain, Psychological Review65, 386 (1958)

    1958

  7. [15]

    Fukushima, Neocognitron: A self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in position, Biological Cybernetics36, 193 (1980)

    K. Fukushima, Neocognitron: A self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in position, Biological Cybernetics36, 193 (1980)

    1980

  8. [16]

    K. He, X. Zhang, S. Ren, and J. Sun, Deep residual learning for image recognition, inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)(2016) pp. 770–778

    2016

  9. [17]

    Krizhevsky,Learning Multiple Layers of Features from Tiny Images, Tech

    A. Krizhevsky,Learning Multiple Layers of Features from Tiny Images, Tech. Rep. (University of Toronto, Toronto, Canada, 2009)

    2009

  10. [18]

    J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei, Imagenet: A large-scale hierarchical image database, in2009 IEEE Conference on Computer Vision and Pattern Recognition (IEEE, 2009) pp. 248–255

    2009

  11. [19]

    Barab ´asi,Network Science(Cambridge University Press, Cambridge, UK, 2016)

    A.-L. Barab ´asi,Network Science(Cambridge University Press, Cambridge, UK, 2016)

    2016

  12. [20]

    M. E. J. Newman,Networks(Oxford University Press, Oxford, United Kingdom, 2018)

    2018

  13. [21]

    Menczer, S

    F. Menczer, S. Fortunato, and C. A. Davis,A first course in net- work science(Cambridge University Press, Cambridge, Eng- land, 2020)

    2020

  14. [22]

    Erd ˝os and A

    P. Erd ˝os and A. R ´enyi, On random graphs I, Publicationes Mathematicae6, 290 (1959)

    1959

  15. [23]

    K.-I. Goh, B. Kahng, and D. Kim, Universal behavior of load distribution in scale-free networks, Phys. Rev. Lett.87, 270701 (2001)

    2001

  16. [24]

    D. E. Rumelhart, G. E. Hinton, and R. J. Williams, Learning representations by back-propagating errors, Nature323, 533 (1986)

    1986

  17. [25]

    Mehta, M

    P. Mehta, M. Bukov, C. H. Wang, A. G. Day, C. Richardson, C. K. Fisher, and D. J. Schwab, A high-bias, low-variance in- troduction to machine learning for physicists, Physics Reports 810, 1 (2019)

    2019

  18. [26]

    Arya, relational graph web,https://github.com/ yasharyaa/relational_graph_web.git(2025), accessed: 2025-07-12

    Y . Arya, relational graph web,https://github.com/ yasharyaa/relational_graph_web.git(2025), accessed: 2025-07-12

    2025

  19. [27]

    Lancichinetti and S

    A. Lancichinetti and S. Fortunato, Community detection al- gorithms: A comparative analysis, Phys. Rev. E80, 056117 (2009)

    2009

  20. [28]

    Arya, Simplified LFR Benchmark Graph,https: //github.com/yasharyaa/Simplified_LFR_Benchmark_ Graph.git(2025), accessed: 2025-07-12

    Y . Arya, Simplified LFR Benchmark Graph,https: //github.com/yasharyaa/Simplified_LFR_Benchmark_ Graph.git(2025), accessed: 2025-07-12

    2025

  21. [29]

    Brenner, The genetics of caenorhabditis elegans, Genetics 77, 71 (1974)

    S. Brenner, The genetics of caenorhabditis elegans, Genetics 77, 71 (1974)

    1974

  22. [30]

    Kaiser and C

    M. Kaiser and C. C. Hilgetag, Nonoptimal component place- ment, but short processing paths, due to long-distance projec- tions in neural systems, PLoS Comput. Biol.2, e95 (2006)

    2006

  23. [31]

    This approach ensures the connectivity of the entire network, preventing the isolation of message exchanges and facilitating effective information flow across the network

    In practice, the actual number of nodes in each community can be smaller than the desired numbers, as we only consider the largest component in the percolation sense for each community. This approach ensures the connectivity of the entire network, preventing the isolation of m...

  24. [32]

    S. Xie, A. Kirillov, R. Girshick, and K. He, Exploring randomly wired neural networks for image recognition, inProceedings of the IEEE/CVF International Conference on Computer Vision (2019) pp. 1284–1293

    2019

  25. [33]

    V . N. Vapnik,The Nature of Statistical Learning Theory (Springer, New York, 1995)

    1995

  26. [34]

    T. K. Rusch, M. M. Bronstein, and S. Mishra, A survey on over- smoothing in graph neural networks, ArXivabs/2303.10993 (2023)

    2023 arXiv

  27. [35]

    Belkin, P

    M. Belkin, P. Niyogi, and V . Sindhwani, Manifold regulariza- tion: A geometric framework for learning from labeled and un- labeled examples, J. Mach. Learn. Res.7, 2399 (2006)

    2006

  28. [36]

    T. N. Kipf and M. Welling, Semi-supervised classification with graph convolutional networks, inInternational Conference on Learning Representations(2017)

    2017

  29. [37]

    Z. Wu, S. Pan, F. Chen, G. Long, C. Zhang, and P. S. Yu, A comprehensive survey on graph neural networks, IEEE transac- tions on neural networks and learning systems32, 4 (2020)

    2020

  30. [38]

    Sporns, G

    O. Sporns, G. Tononi, and G. M. Edelman, Theoretical neu- roanatomy: Relating anatomical and functional connectivity in graphs and cortical connection matrices, Cerebral Cortex10, 127 (2000)

    2000

  31. [39]

    R. Ying, J. You, C. Morris, X. Ren, W. L. Hamilton, and J. Leskovec, Hierarchical graph representation learning with differentiable pooling, inProceedings of the 32nd Interna- tional Conference on Neural Information Processing Systems, NIPS’18 (Curran Associates Inc., Red Hook...

    2018

  32. [40]

    Gilmer, S

    J. Gilmer, S. S. Schoenholz, P. F. Riley, O. Vinyals, and G. E. Dahl, Neural message passing for quantum chemistry, inPro- ceedings of the 34th International Conference on Machine Learning, Proceedings of Machine Learning Research, V ol. 70, edited by D. Precup and Y . W. Teh ...

    2017

  33. [41]

    Tibshirani, Regression shrinkage and selection via the lasso, Journal of the Royal Statistical Society: Series B58, 267 (1996)

    R. Tibshirani, Regression shrinkage and selection via the lasso, Journal of the Royal Statistical Society: Series B58, 267 (1996)

    1996

  34. [42]

    S. Han, J. Pool, J. Tran, and W. Dally, Learning both weights and connections for efficient neural networks, Advances in Neu- ral Information Processing Systems28(2015)

    2015

  35. [43]

    Q. Li, Z. Han, and X.-M. Wu, Deeper insights into graph convo- lutional networks for semi-supervised learning, inProceedings of the AAAI conference on artificial intelligence, V ol. 32 (2018)

    2018

  36. [44]

    Oono and T

    K. Oono and T. Suzuki, Graph neural networks expo- nentially lose expressive power for node classification, arXiv:1905.10947 (2019)

    2019 arXiv

  37. [45]

    M. A. Bertolero, B. T. Yeo, and M. D’Esposito, The modu- lar and integrative functional architecture of the human brain, Proceedings of the National Academy of Sciences112, E6798 (2015)

    2015

  38. [46]

    Frankle and M

    J. Frankle and M. Carbin, The lottery ticket hypothesis: Finding sparse, trainable neural networks, inInternational Conference on Learning Representations(2019)

    2019

  39. [47]

    C. Liu, Y . Han, H. Xu, S. Yang, K. Wang, and Y . Su, A commu- nity detection and graph-neural-network-based link prediction approach for scientific literature, Mathematics12, 369 (2024)

    2024

  40. [48]

    Graph Structure of Neural Networks

    Meta (formerly Facebook) Research, Code for the paper “Graph Structure of Neural Networks”,https://github. com/facebookresearch/graph2nn, accessed: 2025-07-12

    2025

  41. [49]

    S. Son, E. J. Choi, and S. H. Lee, Revisiting small-world net- work models: exploring technical realizations and the equiv- alence of the Newman–Watts and Harary models, J. Korean Phys. Soc.83, 879 (2023)

    2023

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.