Pith. sign in

REVIEW 4 major objections 4 minor 46 references

Assessing the Robustness and Reducibility of Multiplex Networks with Embedding-Aided Interlayer Similarities

T0 review · 4 major / 4 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 new interlayer similarity measure, EATSim, predicts multiplex network robustness and identifies reducible layers better than existing metrics.

desk verdict EATSim is a plausible new embedding-based interlayer similarity metric, but the headline robustness correlations are in-sample: the balance weight ω was tuned on the same robustness labels used for validation, so the claimed advantage over NMI is not yet established. read the letter →

arxiv 2505.06998 v1 pith:7F53EQVX submitted 2025-05-11 cs.SI physics.soc-ph

Haoran Nan , Senquan Wang , Chun Ouyang , Yanchen Zhou , Weiwei Gu This is my paper · ORCID
classification cs.SIphysics.soc-ph
keywords interlayersimilaritymultiplexnetworksnetworkembeddingnode2vecrobustnessreducibilitygeometriccorrelations
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 proposes EATSim, a pairwise interlayer similarity measure for multiplex networks built from node embeddings. EATSim combines two geometric losses: the Pairwise Euclidean Distance (PED) loss, which compares intralayer node-pair distances, and the Aligned Euclidean Distance (AED) loss, which aligns embedding spaces by an orthogonal transform anchored on cross-layer nodes. The authors claim that EATSim captures both local and global structural similarity, and they test it on two downstream tasks. On multiplex robustness prediction, EATSim achieves Pearson correlation 0.882 with the robustness indicator Ω on synthetic networks and 0.856 on 17 real multiplex networks, beating four existing similarity measures. On layer reduction, EATSim selects the most similar layer pairs for aggregation and reaches the highest distinguishability q in several genetic, social, transport, and neuronal networks.

What carries the argument

The central object is the EATSim dissimilarity $D(G^{(\alpha)},G^{(\beta)}) = \omega L_\mathrm{PED} + (1-\omega) L_\mathrm{AED}$, with $\omega=0.5$. The PED loss is the average absolute difference of Euclidean distances between all node pairs' embedding vectors in two layers, capturing local topological similarity; the AED loss is the mean Euclidean distance between aligned anchor-node embeddings after an orthogonal transformation $W$ is found by SVD of $X^{(\alpha)\top}X^{(\beta)}$, capturing global alignment consistency. The paper uses node2vec embeddings with dimension $d=32$, and argues that fixing the random seed makes structurally identical layers yield zero loss. EATSim equals $1-D$, so higher values mark more similar layers.

What would settle it

Evaluate EATSim on a held-out set of real multiplex networks (e.g., ten networks not used in this paper) with ω fixed at 0.5, and compare the Pearson correlation of EATSim with Ω against the NMI baseline; if the correlation no longer exceeds that of NMI, the state-of-the-art robustness claim does not generalize.

Watch

Extended reading notes

Core claim

The central claim is that interlayer similarity of multiplex networks can be faithfully quantified by combining two complementary geometric signals derived from node2vec embeddings: intralayer pairwise Euclidean distances (PED loss) and cross-layer Euclidean distances after an orthogonal alignment (AED loss). The resulting measure, EATSim, is claimed to encode both the local topology of each layer and the global alignment consistency of anchor nodes across layers. The paper argues that this combination is what lets EATSim predict the robustness of a multiplex network against targeted attacks—measured by the size-independent robustness indicator Ω—more accurately than degree-based, entropy-based, mutual-information-based, and edge-overlap baselines, and to identify which layers are safe to aggregate in network reduction. In the paper's own reading, EATSim's success is evidence that embedding geometry, not just static graph statistics, carries the information needed to assess interdependent network resilience and redundancy.

Load-bearing premise

The single balancing weight ω=0.5 for the two losses is fixed globally and was chosen by maximizing the same robustness correlation used to judge the method, so the claim depends on that fitted weight transferring to new networks.

Editorial extensions

If this is right

  • If EATSim is a reliable robustness predictor, then measuring only two layers' embedding geometry is enough to rank which multiplex networks are likely to survive targeted attacks without simulating cascading failures.
  • EATSim-based hierarchical aggregation can shrink a multiplex to fewer layers while preserving distinguishability; in the Candida network it achieves q=0.711 at four layers, the best among compared methods.
  • Because EATSim captures both angular and radial geometric correlations in GMM synthetic networks, it extends robustness prediction beyond angular-only measures like those based on [12].
  • The same similarity score serves both robustness prediction and layer reduction, suggesting a single geometric measure may replace task-specific interlayer similarity indices.

Reading between the lines

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

  • A direct extension is to test whether EATSim's predictive power survives when the base embeddings come from deeper models (e.g., graph neural networks) rather than node2vec, which would show the geometric losses are the active ingredient.
  • Because the two losses are scale-sensitive, applying EATSim to weighted multiplex networks would require a normalization step; exploring that could generalize the measure to transportation and biological networks with heterogeneous edge weights.
  • The reducibility application suggests a greedy algorithm that always merges the pair with highest EATSim; an alternative is to use EATSim as a distance for global clusterings of layers, which might find better non-greedy reductions.
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

4 major / 4 minor

Summary. The manuscript proposes EATSim, an interlayer similarity measure for multiplex networks that combines a Pairwise Euclidean Distance (PED) loss on node2vec embeddings with an Aligned Euclidean Distance (AED) loss after an orthogonal transformation. The method is validated on synthetic networks (BA rewiring, GMM angular/radial correlations) and on 17 real multiplex networks for two downstream tasks: predicting robustness Ω (defined via GMCC size under targeted attack relative to reshuffled networks) and structural reducibility q. The headline results are Pearson correlations PCC=0.882 (synthetic) and 0.856 (real) between EATSim and Ω, which outperform JSD, NMI, D-measure, and LSim, and competitive layer-reduction dendrograms on genetic, social, transport, and neuronal multiplexes.

Significance. The proposed combination of local geometric distance comparisons and cross-layer alignment is sensible, computationally light, and the authors demonstrate monotone behavior against rewiring probability and GMM angular/radial correlations. If the robustness correlations were shown to be predictive rather than in-sample, the method would be a practically useful addition to the interlayer-similarity toolbox. However, the central quantitative claim is weakened by the fact that the only tunable parameter, ω, is selected by maximizing the Pearson correlation with the very robustness measure used for validation. The manuscript also lacks code, confidence intervals, and a clear validation split, so the claimed margin over NMI cannot currently be assessed as predictive.

major comments (4)
  1. [Sec. 2.4 and Sec. 3.2] The hyperparameter ω in Eq. (5) is chosen by scanning values from 0 to 1 and retaining the value that maximizes the Pearson correlation with the robustness indicator Ω, and the same Ω values are then used to report PCC=0.882 and 0.856 as evidence that EATSim 'predicts' robustness. This is in-sample model selection rather than predictive validation. Because the margin over the NMI baseline (PCC=0.819 on real networks) is small, a held-out evaluation (e.g., selecting ω on a subset of GMM configurations or on synthetic networks only and then evaluating on real networks) and bootstrap confidence intervals are necessary to support the state-of-the-art claim.
  2. [Eq. (5)] The PED and AED losses are not normalized or rescaled relative to each other, so ω=0.5 is not a neutral balance but a calibration of two quantities with different units and scales. The transferability of a single global ω to new networks is therefore questionable; the authors should report the relative magnitudes of LPED and LAED, and ideally normalize each term (e.g., z-score or min-max) before combining, or estimate the scale ratio and show that ω=0.5 is robust across networks.
  3. [Abstract and Sec. 3.3] The statement that EATSim achieves 'state-of-the-art performance' in network reducibility is not supported by the reported results: on the London network NMI reaches q=0.521 while EATSim reaches 0.487, and on several datasets all methods perform comparably. The authors should either qualify the claim to specific datasets or regimes or provide an aggregate comparison (e.g., mean rank, number of wins) across all nine reducibility datasets.
  4. [Sec. 3.2 and Fig. 3(a)] The synthetic PCC of 0.882 is computed by pooling GMM networks of sizes 2,000–5,000. Since Ω is designed to be size-independent but EATSim may have residual size dependence, the pooled correlation could be confounded by network size. Report per-size correlations or include size as a covariate in the analysis.
minor comments (4)
  1. [Eq. (4)] The summation runs from k=0 to N, but node indices are 1..N; this should be k=1 to N.
  2. [Sec. 2.4] There is a typo: 'we Iet ω = 0.5' should read 'we let ω = 0.5'.
  3. [References and availability] The supplemental material is referenced as [43] with a URL placeholder, and no code or data availability statement is provided; adding these would make the results reproducible.
  4. [Sec. 3.1] The claim that EATSim exhibits 'remarkable generalization capabilities' is based on one set of node2vec hyperparameters (d=32, walks=10, window=10); a short sensitivity analysis for these choices would strengthen the claim.

Circularity Check

1 steps flagged · score 6.0 of 10

Robustness prediction is partially in-sample: ω is tuned against Ω, the same target used for the reported PCC.

  1. fitted input called prediction [Sec. 2.4 (Eq. 5) and Sec. 3.2 (Eq. 7, Fig. 3)]
    "A comparative analysis of ω values ranging from 0 to 1 (see Supplementary Fig. 1 for details [43]) reveals a non-monotonic relationship between ω and the Pearson correlation to the network robustness indicator, with the maximum correlation observed at ω = 0.5. ... The Pearson Correlation Coefficient (PCC) between EATSim and Ω equals 0.882, higher than other network similarity measurements."

    In Eq. (5), EATSim's dissimilarity is D = ω L_PED + (1−ω) L_AED. Section 2.4 selects ω by maximizing the Pearson correlation with 'the network robustness indicator', i.e., the same Ω defined in Eq. (7) that Section 3.2 uses as the target variable. The reported PCC(EATSim, Ω) = 0.882 is therefore the value of the selection objective itself, evaluated on the same synthetic networks used to choose ω; it is a fitted in-sample maximum over the scanned one-parameter family, not an independent predictive correlation. Because the margin over NMI (0.819 on real networks) is offered as evidence of superiority, and no held-out split or nested validation is reported, the robustness-prediction claim partially reduces to the fitting procedure.

full rationale

The EATSim construction from node2vec embeddings is not circular by itself: PED and AED are defined from embedding geometry without reference to robustness, and the reducibility experiments and correlations with GMM's g/v parameters are not reduced to the fitted weight. However, the headline robustness validation in Sec. 3.2 is compromised by the ω-selection step in Sec. 2.4. Since ω is explicitly chosen to maximize the Pearson correlation with the robustness indicator, reporting the resulting PCC as predictive evidence is an in-sample evaluation: the reported quantity is the same objective used during model selection. Concretely, with ω* = argmax_ω ρ(ω L_PED + (1−ω) L_AED, Ω), the reported ρ(EATSim, Ω) = ρ(ω* L_PED + (1−ω*) L_AED, Ω) = max_ω ρ(...), so the 0.882/0.856 correlations are optimized values rather than out-of-sample predictions. The single fitted scalar limits the degree of inflation, and the paper's self-citations (e.g., [29] for embedding dimension, [32] for Euclidean distances in embedding space) are ordinary methodological references, not load-bearing circularity. Score 6 reflects that the central robustness-prediction claim partially reduces to the fitted parameter, while the rest of the method retains independent content.

Assumptions & free parameters 1 free parameters · 5 assumptions · 0 invented entities

The central similarity computation rests on the assumption that node2vec embeddings preserve geometric structure, that orthogonal alignment makes layers comparable, and that omega equals 0.5 is a valid balance. The omega parameter is a free parameter fitted to the robustness outcome. No new physical or conceptual entities are introduced.

free parameters (1)
  • omega (PED/AED balance weight) = 0.5
    Set in Sec 2.4 after scanning omega from 0 to 1 and choosing the value maximizing Pearson correlation with the robustness indicator Omega. This makes the downstream robustness correlations partly fitted.
assumptions (5)
  • domain assumption node2vec embeddings place structurally similar nodes close in Euclidean space
    Sec 2.1 and 2.2 rely on Gu et al. [32] to treat Euclidean distances between embedding vectors as meaningful topological distances.
  • domain assumption The geometric spaces of different layers are comparable via orthogonal transformation
    Sec 2.3 assumes an orthogonal W suffices to align embedding spaces, following Mikolov et al. [33].
  • domain assumption All layers share the same node set with one-to-one interlinks
    Sec 2.1 defines the multiplex network this way, allowing all nodes to act as alignment anchors in Eq. (4).
  • ad hoc to paper omega = 0.5 is a valid global balance across networks
    Eq. (5) introduces omega, and Sec 2.4 fixes it to 0.5 based on robustness correlation; this is a paper-specific tuning choice.
  • domain assumption GMM angular and radial correlations and rewiring probabilities are valid ground-truth similarity benchmarks
    Sec 3.1 uses these as ground truth for validating EATSim; the reliability of GMM as a benchmark is taken from Kleineberg et al. [44].

how reviews work

0 comments
Cite this review

Pith. "Pith review of Assessing the Robustness and Reducibility of Multiplex Networks with Embedding-Aided Interlayer Similarities." pith.science (2026). https://pith.science/paper/7F53EQVX

@misc{pith2026250506998,
  author       = {Pith},
  title        = {Pith review of: Assessing the Robustness and Reducibility of Multiplex Networks with Embedding-Aided Interlayer Similarities},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7F53EQVX}},
  note         = {Machine review of arXiv:2505.06998}
}
read the original abstract

The study of interlayer similarity of multiplex networks helps to understand the intrinsic structure of complex systems, revealing how changes in one layer can propagate and affect others, thus enabling broad implications for transportation, social, and biological systems. Existing algorithms that measure similarity between network layers typically encode only partial information, which limits their effectiveness in capturing the full complexity inherent in multiplex networks. To address this limitation, we propose a novel interlayer similarity measuring approach named Embedding Aided inTerlayer Similarity (EATSim). EATSim concurrently incorporates intralayer structural similarity and cross-layer anchor node alignment consistency, providing a more comprehensive framework for analyzing interconnected systems. Extensive experiments on both synthetic and real-world networks demonstrate that EATSim effectively captures the underlying geometric similarities between interconnected networks, significantly improving the accuracy of interlayer similarity measurement. Moreover, EATSim achieves state-of-the-art performance in two downstream applications: predicting network robustness and network reducibility, showing its great potential in enhancing the understanding and management of complex systems.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 41 canonical work pages

  1. [1]

    Gleeson, Yamir Moreno, and Mason A

    Mikko Kivelä, Alex Arenas, Marc Barthelemy, James P. Gleeson, Yamir Moreno, and Mason A. Porter. Multilayer networks.Journal of Complex Networks, 2(3):203–271, 07 2014

    work page 2014

  2. [2]

    More is different in real-world multilayer networks

    Manlio De Domenico. More is different in real-world multilayer networks. Nature Physics, 19(9):1247–1262, Sep 2023

    work page 2023

  3. [3]

    Buldyrev, H

    Jianxi Gao, Sergey V. Buldyrev, H. Eugene Stanley, and Shlomo Havlin. Networks formed from interdependent networks.Nature Physics, 8(1):40–48, Jan 2012

    work page 2012

  4. [4]

    A survey of link prediction in complex networks.ACM Comput

    Víctor Martínez, Fernando Berzal, and Juan-Carlos Cubero. A survey of link prediction in complex networks.ACM Comput. Surv., 49(4), dec 2016. 15

    work page 2016

  5. [5]

    Porter, Sergio Gómez, and Alex Arenas

    Manlio De Domenico, Albert Solé-Ribalta, Emanuele Cozzo, Mikko Kivelä, Yamir Moreno, Mason A. Porter, Sergio Gómez, and Alex Arenas. Mathematical formula- tion of multilayer networks.Phys. Rev. X, 3:041022, Dec 2013

    work page 2013

  6. [6]

    The science of science: From the perspective of complex systems

    AnZeng, ZhesiShen, JianlinZhou, JinshanWu, YingFan, YouguiWang, andH.Eugene Stanley. The science of science: From the perspective of complex systems. Physics Reports, 714-715:1–73, 2017. The Science of Science: From the Perspective of Complex Systems

    work page 2017

  7. [7]

    Hanchen Wang, Tianfan Fu, Yuanqi Du, Wenhao Gao, Kexin Huang, Ziming Liu, Payal Chandak, Shengchao Liu, Peter Van Katwyk, Andreea Deac, Anima Anandkumar, Karianne Bergen, Carla P. Gomes, Shirley Ho, Pushmeet Kohli, Joan Lasenby, Jure Leskovec, Tie-Yan Liu, Arjun Manrai, Debora Marks, Bharath Ramsundar, Le Song, Jimeng Sun, Jian Tang, Petar Veličković, Max...

    work page 2023

  8. [8]

    Heterogeneous graph attention network

    Xiao Wang, Houye Ji, Chuan Shi, Bai Wang, Yanfang Ye, Peng Cui, and Philip S Yu. Heterogeneous graph attention network. InThe world wide web conference, pages 2022–2032, 2019

    work page 2022

Show all 46 references
  1. [9]

    Improving link prediction accuracy of network embedding algorithms via rich node attribute information.Journal of Social Computing, 4(4):326–336, 2023

    Weiwei Gu, Jinqiang Hou, and Weiyi Gu. Improving link prediction accuracy of network embedding algorithms via rich node attribute information.Journal of Social Computing, 4(4):326–336, 2023

    2023

  2. [10]

    Pay attention to weak ties: A heteroge- neous multiplex representation learning framework for link prediction.arXiv preprint arXiv:2406.11904, 2024

    Weiwei Gu, Linbi Lv, Gang Lu, and Ruiqi Li. Pay attention to weak ties: A heteroge- neous multiplex representation learning framework for link prediction.arXiv preprint arXiv:2406.11904, 2024

    2024 arXiv

  3. [11]

    Structural reducibility of multilayer networks.Nature Communications, 6(1):6864, Apr 2015

    Manlio De Domenico, Vincenzo Nicosia, Alexandre Arenas, and Vito Latora. Structural reducibility of multilayer networks.Nature Communications, 6(1):6864, Apr 2015

    2015

  4. [12]

    Ángeles Serrano

    Kaj-Kolja Kleineberg, Lubos Buzna, Fragkiskos Papadopoulos, Marián Boguñá, and M. Ángeles Serrano. Geometric correlations mitigate the extreme vulnerability of mul- tiplex networks against targeted attacks.Phys. Rev. Lett., 118:218301, May 2017

    2017

  5. [13]

    Schieber, Laura Carpi, Albert Díaz-Guilera, Panos M

    Tiago A. Schieber, Laura Carpi, Albert Díaz-Guilera, Panos M. Pardalos, Cristina Masoller, and Martín G. Ravetti. Quantification of network structural dissimilarities. Nature Communications, 8(1):13928, Jan 2017

    2017

  6. [14]

    Zizhen Xu and Shauhrat S. Chopra. Interconnectedness enhances network resilience of multimodal public transportation systems for safe-to-fail urban mobility. Nature Communications, 14(1):4291, Jul 2023

    2023

  7. [15]

    Lifei Wang, Rui Nie, Zhang Zhang, Weiwei Gu, Shuo Wang, Anqi Wang, Jiang Zhang, and Jun Cai. A deep generative framework with embedded vector arithmetic and classi- fier for sample generation, label transfer, and clustering of single-cell data.Cell Reports Methods, 3(8), 2023. 16

    2023

  8. [16]

    Enhancing transport properties in in- terconnected systems without altering their structure.Phys

    Arsham Ghavasieh and Manlio De Domenico. Enhancing transport properties in in- terconnected systems without altering their structure.Phys. Rev. Res., 2:013155, Feb 2020

    2020

  9. [17]

    Link prediction in multiplex networks based on interlayer similarity.Physica A: Statistical Mechanics and its Applications, 536:120978, 2019

    Shaghayegh Najari, Mostafa Salehi, Vahid Ranjbar, and Mahdi Jalili. Link prediction in multiplex networks based on interlayer similarity.Physica A: Statistical Mechanics and its Applications, 536:120978, 2019

    2019

  10. [18]

    Zhang and Fred Y

    Ronda J. Zhang and Fred Y. Ye. Measuring similarity for clarifying layer difference in multiplex ad hoc duplex information networks.Journal of Informetrics, 14(1):100987, 2020

    2020

  11. [19]

    Core-based approach to measure pairwiselayersimilarityinmultiplexnetwork

    Debasis Mohapatra, Sourav Kumar Bhoi, Kalyan Kumar Jena, Chittaranjan Mallick, Kshira Sagar Sahoo, NZ Jhanjhi, and Mehedi Masud. Core-based approach to measure pairwiselayersimilarityinmultiplexnetwork. Intelligent Automation & Soft Computing, 34(1):51–64, 2022

    2022

  12. [20]

    A community matching based approach to measuring layer similarity in multilayer networks.Social Networks, 68:1–14, 2022

    Raji Ghawi and Jürgen Pfeffer. A community matching based approach to measuring layer similarity in multilayer networks.Social Networks, 68:1–14, 2022

    2022

  13. [21]

    Characterizing the analogy between hyperbolic embedding and community structure of complex networks.Phys

    Ali Faqeeh, Saeed Osat, and Filippo Radicchi. Characterizing the analogy between hyperbolic embedding and community structure of complex networks.Phys. Rev. Lett., 121:098301, Aug 2018

    2018

  14. [22]

    Comparing alzheimer’s and parkinson’s diseases networks using graph communities structure.BMC Systems Biology, 10(1):25, Mar 2016

    Alberto Calderone, Matteo Formenti, Federica Aprea, Michele Papa, Lilia Alberghina, Anna Maria Colangelo, and Paola Bertolazzi. Comparing alzheimer’s and parkinson’s diseases networks using graph communities structure.BMC Systems Biology, 10(1):25, Mar 2016

    2016

  15. [23]

    Community detection in multi-layer graphs: A survey

    Jungeun Kim and Jae-Gil Lee. Community detection in multi-layer graphs: A survey. SIGMOD Rec., 44(3):37–48, dec 2015

    2015

  16. [24]

    Network mutual information measures for graph similarity.Communications Physics, 7(1):335, 2024

    Helcio Felippe, Federico Battiston, and Alec Kirkley. Network mutual information measures for graph similarity.Communications Physics, 7(1):335, 2024

    2024

  17. [25]

    Hamilton, Rex Ying, and Jure Leskovec

    William L. Hamilton, Rex Ying, and Jure Leskovec. Representation learning on graphs: Methods and applications.IEEE Data Eng. Bull., 40(3):52–74, 2017

    2017

  18. [26]

    A survey on network embedding

    Peng Cui, Xiao Wang, Jian Pei, and Wenwu Zhu. A survey on network embedding. IEEE Transactions on Knowledge and Data Engineering, 31(5):833–852, 2019

    2019

  19. [27]

    Mwtp: A heterogeneous multiplex representation learning framework for link prediction of weak ties.Neural Networks, page 107489, 2025

    Weiwei Gu, Linbi Lv, Gang Lu, and Ruiqi Li. Mwtp: A heterogeneous multiplex representation learning framework for link prediction of weak ties.Neural Networks, page 107489, 2025

    2025

  20. [28]

    Discovering latent node informa- tion by graph attention network.Scientific reports, 11(1):6967, 2021

    Weiwei Gu, Fei Gao, Xiaodan Lou, and Jiang Zhang. Discovering latent node informa- tion by graph attention network.Scientific reports, 11(1):6967, 2021. 17

    2021

  21. [29]

    Principled approach to the selection of the embedding dimension of networks

    Weiwei Gu, Aditya Tandon, Yong-Yeol Ahn, and Filippo Radicchi. Principled approach to the selection of the embedding dimension of networks. Nature Communications, 12(1):3772, 2021

    2021

  22. [30]

    Graph neural networks in node classification: survey and evaluation.Mach

    Shunxin Xiao, Shiping Wang, Yuanfei Dai, and Wenzhong Guo. Graph neural networks in node classification: survey and evaluation.Mach. Vision Appl., 33(1), jan 2022

    2022

  23. [31]

    A general deep reinforcement learning hyperheuristic framework for solving combinatorial opti- mization problems

    Jakob Kallestad, Ramin Hasibi, Ahmad Hemmati, and Kenneth Sörensen. A general deep reinforcement learning hyperheuristic framework for solving combinatorial opti- mization problems. European Journal of Operational Research, 309(1):446–468, 2023

    2023

  24. [32]

    The hidden flow structure and metric space of network embedding algorithms based on random walks.Scientific re- ports, 7(1):13114, 2017

    Weiwei Gu, Li Gong, Xiaodan Lou, and Jiang Zhang. The hidden flow structure and metric space of network embedding algorithms based on random walks.Scientific re- ports, 7(1):13114, 2017

    2017

  25. [33]

    Exploitingsimilaritiesamonglanguages for machine translation, 2013

    TomasMikolov, QuocV.Le, andIlyaSutskever. Exploitingsimilaritiesamonglanguages for machine translation, 2013

    2013

  26. [34]

    Catastrophic cascade of failures in interdependent networks.Nature, 464(7291):1025– 1028, Apr 2010

    SergeyV.Buldyrev, RoniParshani, GeraldPaul, H.EugeneStanley, andShlomoHavlin. Catastrophic cascade of failures in interdependent networks.Nature, 464(7291):1025– 1028, Apr 2010

    2010

  27. [35]

    Structural measures for multi- plex networks.Phys

    Federico Battiston, Vincenzo Nicosia, and Vito Latora. Structural measures for multi- plex networks.Phys. Rev. E, 89:032804, Mar 2014

    2014

  28. [36]

    node2vec: Scalable feature learning for networks

    Aditya Grover and Jure Leskovec. node2vec: Scalable feature learning for networks. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discov- ery and Data Mining, KDD ’16, page 855–864, New York, NY, USA, 2016. Association for Computing Machinery

    2016

  29. [37]

    Line: Large-scale information network embedding

    Jian Tang, Meng Qu, Mingzhe Wang, Ming Zhang, Jun Yan, and Qiaozhu Mei. Line: Large-scale information network embedding. In Proceedings of the 24th international conference on world wide web, pages 1067–1077, 2015

    2015

  30. [38]

    Inductive representation learning on large graphs

    Will Hamilton, Zhitao Ying, and Jure Leskovec. Inductive representation learning on large graphs. Advances in neural information processing systems, 30, 2017

    2017

  31. [39]

    Semi-supervised classification with graph convolu- tional networks.arXiv preprint arXiv:1609.02907, 2016

    Thomas N Kipf and Max Welling. Semi-supervised classification with graph convolu- tional networks.arXiv preprint arXiv:1609.02907, 2016

    2016 arXiv

  32. [40]

    Efficient estimation of word representations in vector space.arXiv preprint arXiv:1301.3781, 2013

    Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. Efficient estimation of word representations in vector space.arXiv preprint arXiv:1301.3781, 2013

    2013 arXiv

  33. [41]

    Network communitydetectionvianeuralembeddings

    Sadamori Kojaku, Filippo Radicchi, Yong-Yeol Ahn, and Santo Fortunato. Network communitydetectionvianeuralembeddings. Nature Communications, 15(1):9446, 2024

    2024

  34. [42]

    Umap: Uniform manifold approxi- mation and projection for dimension reduction.arXiv preprint arXiv:1802.03426, 2018

    Leland McInnes, John Healy, and James Melville. Umap: Uniform manifold approxi- mation and projection for dimension reduction.arXiv preprint arXiv:1802.03426, 2018. 18

    2018 arXiv

  35. [43]

    SeeSupplementalMaterialat[urlwillbeinsertedbypublisher]fordetailedexperimental analysis of parameterω selection; investigation of the relationship between EATSim and geometric correlations in GMM; exploration of how clustering coefficient and tempera- ture T affect network rob...

  36. [44]

    Ángeles Serrano, and Fragkiskos Papadopou- los

    Kaj-Kolja Kleineberg, Marián Boguñá, M. Ángeles Serrano, and Fragkiskos Papadopou- los. Hidden geometric correlations in real multiplex networks. Nature Physics, 12(11):1076–1081, Nov 2016

    2016

  37. [45]

    The robustness of interdependent directed networks with intra-layer angular correlations.Frontiers in Physics, 9:755567, 2021

    Zongning Wu, Zengru Di, and Ying Fan. The robustness of interdependent directed networks with intra-layer angular correlations.Frontiers in Physics, 9:755567, 2021

    2021

  38. [46]

    Saulo D. S. Reis, Yanqing Hu, Andrés Babino, José S. Andrade Jr, Santiago Canals, Mariano Sigman, and Hernán A. Makse. Avoiding catastrophic failure in correlated networks of networks.Nature Physics, 10(10):762–767, Oct 2014. 19

    2014

Pith tools

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