Pith. sign in

REVIEW 4 major objections 5 minor 53 references

Network classification through random walks

T0 review · 4 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read This paper claims that statistics of node visits and walk lengths from self-avoiding walks, combined with traditional random-walk baselines, form a feature set that classifies real-world networks more accurately than four established…

desk verdict Sensible, simply described feature set, but the headline accuracy claim is contradicted by the paper's own tables. read the letter →

arxiv 2505.21706 v1 pith:W6CU3WXL submitted 2025-05-27 cs.SI cs.LGphysics.soc-ph

Gonzalo Travieso , Joao Merenda , Odemir M. Bruno This is my paper
classification cs.SIcs.LGphysics.soc-ph
keywords networkclassificationrandomwalksself-avoidingfeatureextractionlineardiscriminantanalysisgraphembeddingscomplexnetworksnoiserobustness
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 paper tries to establish that simple statistics extracted from random walks—specifically, how often each node is visited by self-avoiding walks, and how long those walks run—work as stable, size-invariant signatures for classifying whole networks by the type of system that produced them. If true, this matters because network classification currently relies on heavier machinery such as graph embeddings, automata, or hand-picked structural metrics, and a cheap walk-based feature vector could give comparable or better accuracy with less tuning. The authors report that the self-avoiding-walk visit and length features beat four established baselines on nine of twelve datasets and degrade less under edge rewiring noise.

What carries the argument

The machinery is the self-avoiding walk (SAW) and its limited-memory variant, used as probes of network structure. For each node, the method launches 10 walkers, records how many times each node is visited (normalized by total walkers), and for SAWs also records walk lengths; from these it forms features such as mean, standard deviation, quartiles, skewness, kurtosis, and entropy of visit counts, and the same statistics of SAW lengths. The key comparison object is the difference $s_i - t_i$ between SAW visits and traditional random-walk visits (and $r_i^{(m)} - t_i$ for memory-limited walks), which isolates the structural information that self-avoidance extracts. These features are fed to a linear discriminant classifier under 10-fold cross-validation.

What would settle it

Re-run the experiment with 1,000 walkers per node on the same datasets and check whether accuracies stay within the reported standard deviations; if they shift materially, the 10-walker choice is not stable, and separately re-tuning the baselines (e.g., the graph2vec embedding dimension or the DTW memory set) could reveal whether the 9-of-12 advantage persists under better-tuned competition.

Watch

Extended reading notes

Core claim

The central discovery is that the combination of two statistics from self-avoiding walks—node-visit counts and walk-length distributions—plus the corresponding traditional random-walk visit counts, yields a feature set that outperforms structural measures, life-like network automata, deterministic tourist walks, and graph2vec on 9 of 12 datasets, ties on one, and loses on two (Firmicutes-Bacillus and Collab). The same feature set keeps its accuracy better under edge rewiring: its average drop is 24.73%, compared with 26.13% for DTW, 28.80% for LLNA, 45.20% for structural measures, and 52.10% for graph2vec.

Load-bearing premise

The comparison assumes that the baseline methods were run with fairly chosen, representative settings, and that the walk statistics gathered from just 10 walkers per node are stable enough to stand for the whole network.

Editorial extensions

If this is right

  • If the claim holds, network classification no longer requires trained graph embeddings or automata; cheap walk statistics are enough for competitive accuracy on many datasets.
  • Memory-limited walks add little beyond the self-avoiding walk alone, so practitioners can skip them and save the extra computation.
  • The 24.73% noise drop suggests walk-based features are a robust choice for noisy or partially observed networks.
  • The method's failures on Firmicutes-Bacillus and Collab point to specific structures that remain hard, and the paper's suggestion of first-visit times offers a concrete extension to try.
  • Since walkers and steps both scale with network size, the method is limited on very large networks; a fixed walk-length cap is the paper's proposed, still untested, remedy.

Reading between the lines

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

  • Editorial inference: The difference $s_i - t_i$ may be interpretable as how strongly each node's neighborhood traps or frees a self-avoiding walker, which could connect to local properties like dead-end density or tree-likeness that the paper does not analyze.
  • Editorial inference: Because the features use no node attributes, the same method could be applied to directed or weighted networks simply by changing the transition rule; the paper tests unweighted undirected graphs only.
  • Editorial inference: The strong performance under rewiring suggests visit statistics encode large-scale mixing rather than exact edge positions, so the features may transfer across networks of the same model but different sizes—a test the paper does not explicitly run.
  • Editorial inference: A natural sensitivity check would be to vary the number of walkers $W$ and confirm that accuracies plateau; the paper uses $W=10$ without a convergence study.
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

4 major / 5 minor

Summary. The paper proposes a network classification method based on statistics extracted from traditional random walks, self-avoiding walks, and limited-memory self-avoiding walks. Features are derived from normalized node-visit counts and SAW walk lengths, then fed to an LDA classifier with 10-fold cross-validation. The method is evaluated on 12 datasets spanning synthetic, metabolic, bioinformatics, and social networks, and compared with structural measures, LLNA, deterministic tourist walks, and Graph2vec. The authors report that the SAW length+visits feature set achieves the highest accuracy in 9 of 12 datasets and degrades less under noise than the baselines. The proposed feature extraction is simple and clearly described, but the key comparison table is internally inconsistent with the detailed results, and the feature-selection protocol is not specified.

Significance. If the reported results are correct, this would be a competitive and lightweight alternative to deep graph embeddings for network classification, with a useful robustness property under edge noise. The paper has several strengths: the feature extraction procedure is easy to reproduce in principle, the evaluation spans a wide range of datasets, and the comparison includes four diverse baselines. However, the central empirical claim rests on Table 4, which appears to contain a systematic data-entry error: the 'Random walks' column values for Synthetic, Proteins, Collab, and IMDB-Multi are exactly shifted relative to the SAW length+visits row in Table 3. Until this is corrected and the feature-selection protocol is clarified, the headline accuracy and noise-robustness claims are not supported.

major comments (4)
  1. [Table 4 vs Tables 1–3] The 'Random walks' column of Table 4 is not consistent with the detailed results in Tables 1–3. For the SAW length+visits feature set, Table 3 reports Synthetic=78.1, Proteins=65.1, Collab=49.3, and IMDB-Multi=100.0; Table 4 reports for the same ordering Synthetic=100.0, Proteins=78.1, Collab=65.1, and IMDB-Multi=49.3. This is exactly a one-dataset shift for the last three entries, and the Synthetic value 100.0 appears nowhere in Table 3 (whose maximum for Synthetic is 78.43). Therefore the headline claim in Section 4 of highest accuracy in 9 of 12 datasets, and the noise-robustness comparison built on the same method, are not currently supported by the reported numbers. If Table 4 was intended to report a per-dataset best combination rather than a single fixed feature set, that selection must be disclosed, and the comparison against fixed baselines would need to be interpreted accordingly.
  2. [Section 4, feature-set selection] The manuscript does not state how the feature set reported in Table 4 (apparently 'SAW length+visits') was selected. Section 4 describes evaluating multiple combinations and reports which were best ('SAW visits set alone outperformed the others... in 11 of the 12 datasets', 'SAW lengths and visits... achieving the best results in 8 out of 12 datasets'). If these evaluations used the test folds, then choosing the best-performing feature set on the test data introduces selection bias, and the comparison against fixed baselines is not a fair estimate of performance. Please specify a selection protocol that uses only training folds or nested cross-validation.
  3. [Sections 3.1.1–3.1.4 and Table 4] The baseline methods are not described with the same implementation detail as the proposed method. For structural measures, the feature list is given but no normalization or implementation choices are described; for LLNA, the transition rule is given but the number of time steps and descriptor parameters are not; for Graph2vec, no hyperparameters (embedding dimension, number of epochs, WL iterations) are reported, and Table 4 shows standard deviations above 20 percentage points for Animal, Enzymes, and Proteins. Without these details, the reader cannot judge whether the baselines are representative, and the claimed improvements may reflect unfavorable baseline settings rather than the merits of the proposed features.
  4. [Section 3.4] The experimental setup fixes W=10 walkers per node and S=N steps without any convergence or stability analysis. The reported standard deviations in Tables 1–3 are large for several datasets (e.g., Protist entries around 10–34 percentage points), and for datasets with very few samples per class (Protist has 5 per class, Plant has 9 per class), 10-fold cross-validation folds contain only 2–3 samples, making LDA estimates unstable. Please provide evidence that the results are insensitive to W and S, and use a validation scheme suitable for small sample sizes, such as repeated cross-validation.
minor comments (5)
  1. [Sections 4 and 5] Section 4 states that the proposed method achieves highest accuracy in 9 of 12 datasets, while Section 5 states 10 of 12; these counts must be reconciled.
  2. [Table 1] Table 1 reports identical accuracy and standard deviation, 96.4 (5.4), for both Actinobacteria and Animals under 'SAW length+visits', and many other entries share the same standard deviation 5.4; please verify that these values were not duplicated.
  3. [Figure 2 caption] The caption of Figure 2 refers to 'sas' and 'sav' feature sets, but these acronyms are not defined in the text.
  4. [Table 4, Enzymes row] Table 4 reports 0.0% accuracy for structural measures and DTW on Enzymes; if this is correct, please explain why, as a zero accuracy on a six-class problem is surprising and may indicate a degenerate LDA fit.
  5. [Table 4 column header] The column name 'Random walks' is potentially confusing because the proposed method includes self-avoiding walks; consider renaming it 'Proposed (SAW length+visits)' for clarity.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the proposed features are measured on held-out folds and are not derived from the classification targets or from load-bearing self-citations.

full rationale

The paper is an empirical feature-extraction and benchmark study. Features are computed from network structure and random-walk statistics (visits, walk lengths, limited-memory walks) without using class labels or baseline accuracies; classification performance is obtained with LDA under 10-fold cross-validation on held-out folds. No predicted quantity is defined in terms of an input in a way that would make the result true by construction. Self-citations appear in descriptions of background and baseline methods (e.g., DTW configuration [45,46], LLNA [7,8], random-walk exploration [21]), but they are not used as uniqueness theorems or as the justification that the proposed features are correct, so they are not load-bearing for the central claim. The apparent shift between Table 3's 'SAW length+visits' values and Table 4's 'Random walks' column for Proteins, Collab, IMDB-Multi, and Synthetic is a data-consistency/correctness concern that should be checked by the authors; it does not constitute a circular derivation, because it does not show that any result is equivalent to its own input by construction. Consequently, no circularity step is identified, and the appropriate score is 0.

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

The paper introduces no new theoretical entities. It relies on several empirical and methodological assumptions: that random walk statistics capture class structure, that the chosen hyperparameters (W=10, S=N) are sufficient, and that the baselines are fairly compared. The free parameters are all method hyperparameters, not fitted to optimize test accuracy, which limits circularity concerns. The main risk is that the empirical results may be sensitive to these choices.

free parameters (5)
  • Number of walkers W = 10
    Chosen without justification or convergence analysis. Results could change with more walkers.
  • Maximum walk length S = S = N (number of nodes)
    Chosen as a limit for traditional and memory-limited walks. Not justified as sufficient for convergence.
  • Memory sizes m for LMW = m = 1 to 10
    Range explored empirically; Figure 1 shows saturation around m = 7 or 8, so the range is a scanned parameter.
  • DTW baseline parameters = mu = 1, 2; min and max rules; m = 5
    Taken from prior work, but no tuning or sensitivity analysis is reported for this baseline.
  • LLNA rule = B01678-S0457
    A single rule is chosen from the life-like automaton rule space; no justification or search is reported.
assumptions (4)
  • domain assumption Statistics of random walk visit counts and walk lengths are informative features for network classification.
    This is the central hypothesis of the paper, tested empirically but not derived.
  • domain assumption LDA with 10-fold cross-validation provides reliable accuracy estimates on the given datasets.
    Standard practice, but questionable for tiny datasets like Protist (5 samples per class) and Plant (9 per class).
  • domain assumption The baseline feature extraction methods (structural, LLNA, DTW, Graph2vec) are implemented faithfully and fairly.
    The baselines are cited but not described in implementation detail, and Graph2vec has very high standard deviations.
  • domain assumption The datasets accurately represent distinct network classes.
    The metabolic and TUDataset classifications are taken as ground truth from the sources.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Network classification through random walks." pith.science (2026). https://pith.science/paper/W6CU3WXL

@misc{pith2026250521706,
  author       = {Pith},
  title        = {Pith review of: Network classification through random walks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/W6CU3WXL}},
  note         = {Machine review of arXiv:2505.21706}
}
read the original abstract

Network models have been widely used to study diverse systems and analyze their dynamic behaviors. Given the structural variability of networks, an intriguing question arises: Can we infer the type of system represented by a network based on its structure? This classification problem involves extracting relevant features from the network. Existing literature has proposed various methods that combine structural measurements and dynamical processes for feature extraction. In this study, we introduce a novel approach to characterize networks using statistics from random walks, which can be particularly informative about network properties. We present the employed statistical metrics and compare their performance on multiple datasets with other state-of-the-art feature extraction methods. Our results demonstrate that the proposed method is effective in many cases, often outperforming existing approaches, although some limitations are observed across certain datasets.

Figures

Figures reproduced from arXiv: 2505.21706 by the authors.

Figure 1
Figure 1. Classification performance of the limited-memory random walk as a function of the memory parameter [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗
Figure 2
Figure 2. Classification performance under increasing noise levels. The proposed Random Walk-based method (using [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

53 extracted references · 51 canonical work pages

  1. [1]

    Costa, Osvaldo N

    Luciano da F. Costa, Osvaldo N. Oliveira Jr., Gonzalo Travieso, Francisco A. Rodrigues, Paulino R. Villas Boas, Lucas Antiqueira, Matheus P. Viana, and Luis E. C. Rocha. Analyzing and modeling real-world phenomena with complex networks: a survey of applications.Advances in Physics, 60(3):329–412, 2011

    work page 2011

  2. [2]

    Costa, Francisco A

    Luciano da F. Costa, Francisco A. Rodrigues, Gonzalo Travieso, and Paulino R. Villas Boas. Characterization of complex networks: A survey of measurements.Advances in physics, 56(1):167–242, 2007

    work page 2007

  3. [3]

    Neiva and Odemir M

    Mariane B. Neiva and Odemir M. Bruno. Exploring ordered patterns in the adjacency matrix for improving machine learning on complex networks.Physica A: Statistical Mechanics and its Applications, 626:129086, 2023

    work page 2023

  4. [4]

    Complex network classification with convolutional neural network

    Ruyue Xin, Jiang Zhang, and Yitong Shao. Complex network classification with convolutional neural network. Tsinghua Science and Technology, 25(4):447–457, 2020

    work page 2020

  5. [5]

    A graph convolutional neural network for classification of building patterns using spatial vector data.ISPRS Journal of Photogrammetry and Remote Sensing, 150:259– 273, 2019

    Xiongfeng Yan, Tinghua Ai, Min Yang, and Hongmei Yin. A graph convolutional neural network for classification of building patterns using spatial vector data.ISPRS Journal of Photogrammetry and Remote Sensing, 150:259– 273, 2019

    work page 2019

  6. [6]

    A deep graph convolutional neural network architecture for graph classification.PLoS ONE, 18(3):e0279604, 2023

    Yuchen Zhou, Hongtao Huo, Zhiwen Hou, and Fanliang Bu. A deep graph convolutional neural network architecture for graph classification.PLoS ONE, 18(3):e0279604, 2023

    work page 2023

  7. [7]

    Gisele H. B. Miranda, Jeaneth Machicao, and Odemir M. Bruno. Exploring spatio-temporal dynamics of cellular automata for pattern recognition in networks.Scientific Reports, 6(1):1–15, 2016

    work page 2016

  8. [8]

    Ribas, Jeaneth Machicao, and Odemir M

    Lucas C. Ribas, Jeaneth Machicao, and Odemir M. Bruno. Life-like network automata descriptor based on binary patterns for network classification.Information Sciences, 515:156–168, 2020

    work page 2020

Show all 53 references
  1. [9]

    Kallil M. C. Zielinski, Lucas C. Ribas, Jeaneth Machicao, and Odemir M. Bruno. A network classification method based on density time evolution patterns extracted from network automata.Pattern Recognition, 146:109946, 2024. 9 Network classification through random walks

    2024

  2. [10]

    Supervised random walks: predicting and recommending links in social networks

    Lars Backstrom and Jure Leskovec. Supervised random walks: predicting and recommending links in social networks. InProceedings of the Fourth ACM International Conference on Web Search and Data Mining, WSDM ’11, pages 635—-644. Association for Computing Machinery, 2011

    2011

  3. [11]

    Silva and Liang Zhao

    Thiago C. Silva and Liang Zhao. High-level pattern-based classification via tourist walks in networks.Information Sciences, 294:109–126, 2015

    2015

  4. [12]

    Gonçalves, André R

    Wesley N. Gonçalves, André R. Backes, Alexandre S. Martinez, and Odemir M. Bruno. Texture descriptor based on partially self-avoiding deterministic walker on networks.Expert Systems with Applications, 39(15):11818–11829, 2012

    2012

  5. [13]

    Gonçalves, Alexandre S

    Wesley N. Gonçalves, Alexandre S. Martinez, and Odemir M. Bruno. Complex network classification using par- tially self-avoiding deterministic walks.Chaos: An Interdisciplinary Journal of Nonlinear Science, 22(3):033139, 2012

    2012

  6. [14]

    B. D. Hughes.Random Walks and Randon Environments, volume V olume 1: Random Walks. Oxford University Press, 1995

    1995

  7. [15]

    John Wiley & Sons, 2013

    Oliver C Ibe.Elements of random walk and diffusion processes. John Wiley & Sons, 2013

    2013

  8. [16]

    Foraging and movement paths of female reindeer: insights from fractal analysis, correlated random walks, and lévy flights.Canadian Journal of Zoology, 80(5):854–865, 2002

    Anders Mårell, John P Ball, and Annika Hofgaard. Foraging and movement paths of female reindeer: insights from fractal analysis, correlated random walks, and lévy flights.Canadian Journal of Zoology, 80(5):854–865, 2002

    2002

  9. [17]

    Random walks: A review of algorithms and applications.IEEE Transactions on Emerging Topics in Computational Intelligence, 4(2):95–107, 2019

    F Xia, J Liu, H Nie, Y Fu, L Wan, and X Kong. Random walks: A review of algorithms and applications.IEEE Transactions on Emerging Topics in Computational Intelligence, 4(2):95–107, 2019

    2019

  10. [18]

    Random-walk model of chain-polymer adsorption at a surface.The Journal of Chemical Physics, 43(7):2392–2407, 1965

    Robert J Rubin. Random-walk model of chain-polymer adsorption at a surface.The Journal of Chemical Physics, 43(7):2392–2407, 1965

    1965

  11. [19]

    Porter, and Renaud Lambiotte

    Naoki Masuda, Mason A. Porter, and Renaud Lambiotte. Random walks and diffusion on networks.Physics Reports, 716–717:1–58, November 2017

    2017

  12. [20]

    Random Walks on Complex Networks.Physical Review Letters, 92(11):118701, March 2004

    Jae Dong Noh and Heiko Rieger. Random Walks on Complex Networks.Physical Review Letters, 92(11):118701, March 2004

    2004

  13. [21]

    Exploring complex networks through random walks.Physical Review E, 75(1):016102, January 2007

    Luciano Da Fontoura Costa and Gonzalo Travieso. Exploring complex networks through random walks.Physical Review E, 75(1):016102, January 2007

    2007

  14. [22]

    Biased random walks in complex networks: The role of local navigation rules.Physical Review E, 80(1):016107, July 2009

    Agata Fronczak and Piotr Fronczak. Biased random walks in complex networks: The role of local navigation rules.Physical Review E, 80(1):016107, July 2009

    2009

  15. [23]

    M.E. J. Newman. A measure of betweenness centrality based on random walks.Social Networks, 27(1):39–54, January 2005

    2005

  16. [24]

    Complex network comparison using random walks

    Shan Lu, Jieqi Kang, Weibo Gong, and Don Towsley. Complex network comparison using random walks. In Proceedings of the 23rd International Conference on World Wide Web, WWW ’14 Companion, pages 727–730, New York, NY , USA, April 2014. Association for Computing Machinery

    2014

  17. [25]

    Random walks on complex trees.Physical Review E, 78(1):011114, July 2008

    Andrea Baronchelli, Michele Catanzaro, and Romualdo Pastor-Satorras. Random walks on complex trees.Physical Review E, 78(1):011114, July 2008

    2008

  18. [26]

    Random Walks and Search in Time-Varying Networks.Physical Review Letters, 109(23):238701, December 2012

    Nicola Perra, Andrea Baronchelli, Delia Mocanu, Bruno Gonçalves, Romualdo Pastor-Satorras, and Alessandro Vespignani. Random Walks and Search in Time-Varying Networks.Physical Review Letters, 109(23):238701, December 2012

    2012

  19. [27]

    Random walks on temporal networks.Physical Review E, 85(5):056115, May 2012

    Michele Starnini, Andrea Baronchelli, Alain Barrat, and Romualdo Pastor-Satorras. Random walks on temporal networks.Physical Review E, 85(5):056115, May 2012

    2012

  20. [28]

    Random walks on weighted networks.Physical Review E, 87(1):012112, January 2013

    Zhongzhi Zhang, Tong Shan, and Guanrong Chen. Random walks on weighted networks.Physical Review E, 87(1):012112, January 2013

    2013

  21. [29]

    Random walks on weighted networks: A survey of local and non-local dynamics

    A P Riascos and José L Mateos. Random walks on weighted networks: A survey of local and non-local dynamics. Journal of Complex Networks, 9(5):cnab032, September 2021

    2021

  22. [30]

    Random walk centrality in interconnected multilayer networks.Physica D: Nonlinear Phenomena, 323–324:73–79, June 2016

    Albert Solé-Ribalta, Manlio De Domenico, Sergio Gómez, and Alex Arenas. Random walk centrality in interconnected multilayer networks.Physica D: Nonlinear Phenomena, 323–324:73–79, June 2016

    2016

  23. [31]

    Random walks on hypergraphs.Physical Review E, 101(2):022308, February 2020

    Timoteo Carletti, Federico Battiston, Giulia Cencetti, and Duccio Fanelli. Random walks on hypergraphs.Physical Review E, 101(2):022308, February 2020

    2020

  24. [32]

    Self-avoiding walks on irregular networks.Journal of Physics A: Mathematical and General, 10(2):197–204, February 1977

    B J Hiley, J L Finney, and T Burke. Self-avoiding walks on irregular networks.Journal of Physics A: Mathematical and General, 10(2):197–204, February 1977. 10 Network classification through random walks

    1977

  25. [33]

    Diffusion on self-avoiding walk networks.Journal of Physics A: Mathematical and General, 22(17):3621–3627, September 1989

    S S Manna, A J Guttmann, and A K Roy. Diffusion on self-avoiding walk networks.Journal of Physics A: Mathematical and General, 22(17):3621–3627, September 1989

    1989

  26. [34]

    Herrero and Martha Saboyá

    Carlos P. Herrero and Martha Saboyá. Self-avoiding walks and connective constants in small-world networks. Physical Review E, 68(2):026106, August 2003

    2003

  27. [35]

    Carlos P. Herrero. Self-avoiding walks on scale-free networks.Physical Review E, 71(1):016103, January 2005

    2005

  28. [36]

    Exploring complex networks by walking on them.Physical Review E, 71(1):016107, January 2005

    Shi-Jie Yang. Exploring complex networks by walking on them.Physical Review E, 71(1):016107, January 2005

    2005

  29. [37]

    López Millán, Vicent Cholvi, Luis López, and Antonio Fernández Anta

    Víctor M. López Millán, Vicent Cholvi, Luis López, and Antonio Fernández Anta. A model of self-avoiding random walks for searching complex networks.Networks, 60(2):71–85, September 2012

    2012

  30. [38]

    Network exploration using true self-avoiding walks.Physical Review E, 94(4):042309, October 2016

    Yup Kim, Seokjong Park, and Soon-Hyung Yook. Network exploration using true self-avoiding walks.Physical Review E, 94(4):042309, October 2016

    2016

  31. [39]

    Characteristic times of biased random walks on complex networks.Physical Review E, 89(1):012803, January 2014

    Moreno Bonaventura, Vincenzo Nicosia, and Vito Latora. Characteristic times of biased random walks on complex networks.Physical Review E, 89(1):012803, January 2014

    2014

  32. [40]

    Maximal-entropy random walks in complex networks with limited information.Physical Review E, 83(3):030103, March 2011

    Roberta Sinatra, Jesús Gómez-Gardeñes, Renaud Lambiotte, Vincenzo Nicosia, and Vito Latora. Maximal-entropy random walks in complex networks with limited information.Physical Review E, 83(3):030103, March 2011

    2011

  33. [41]

    Reactive random walkers on complex networks.Physical Review E, 98(5):52302, 2018

    Giulia Cencetti, Federico Battiston, Duccio Fanelli, and Vito Latora. Reactive random walkers on complex networks.Physical Review E, 98(5):52302, 2018

    2018

  34. [42]

    Hierarchical characterization of complex networks

    Luciano da Fontoura Costa and Filipi Nascimento Silva. Hierarchical characterization of complex networks. Journal of Statistical Physics, 125:841–872, 2006

    2006

  35. [43]

    Campiteli, Pablo D

    Mônica G. Campiteli, Pablo D. Batista, Osame Kinouchi, and Alexandre S. Martinez. Deterministic walks as an algorithm of pattern recognition.Physical Review E—Statistical, Nonlinear, and Soft Matter Physics, 74(2):026703, 2006

    2006

  36. [44]

    Backes, Wesley N

    André R. Backes, Wesley N. Gonçalves, Alexandre S. Martinez, and Odemir M. Bruno. Texture analysis and classification using deterministic tourist walk.Pattern Recognition, 43(3):685–694, 2010

    2010

  37. [45]

    Ribas and Odemir M

    Lucas C. Ribas and Odemir M. Bruno. Fast deterministic tourist walk for texture analysis.arXiv preprint arXiv:1611.08624, 2016

    2016 arXiv

  38. [46]

    João V . B. S. Merenda and Odemir M. Bruno. Using deterministic self-avoiding walks as a small-world metric on watts–strogatz networks.Physica A: Statistical Mechanics and its Applications, 621:128713, 2023

    2023

  39. [47]

    graph2vec: Learning distributed representations of graphs.arXiv preprint arXiv:1707.05005, 2017

    Annamalai Narayanan, Mahinthan Chandramohan, Rajasekar Venkatesan, Lihui Chen, Yang Liu, and Shantanu Jaiswal. graph2vec: Learning distributed representations of graphs.arXiv preprint arXiv:1707.05005, 2017

    2017 arXiv

  40. [48]

    Weisfeiler-lehman graph kernels.Journal of Machine Learning Research, 12(9), 2011

    Nino Shervashidze, Pascal Schweitzer, Erik Jan Van Leeuwen, Kurt Mehlhorn, and Karsten M Borgwardt. Weisfeiler-lehman graph kernels.Journal of Machine Learning Research, 12(9), 2011

    2011

  41. [49]

    Pardalos, and Theodore B

    Petros Xanthopoulos, Panos M. Pardalos, and Theodore B. Trafalis. Linear discriminant analysis.Robust data mining, pages 27–33, 2013

    2013

  42. [50]

    Performance evaluation of classification algorithms by k-fold and leave-one-out cross validation

    Tzu-Tsung Wong. Performance evaluation of classification algorithms by k-fold and leave-one-out cross validation. Pattern recognition, 48(9):2839–2846, 2015

    2015

  43. [51]

    Kegg.Nucleic Acids Research, 28(1):27–30, 2000

    Minoru Kanehisa and Susumu Goto. Kegg.Nucleic Acids Research, 28(1):27–30, 2000

    2000

  44. [52]

    Kegg as a reference resource for gene and protein annotation.Nucleic acids research, 44(D1):D457–D462, 2016

    Minoru Kanehisa, Yoko Sato, Masayuki Kawashima, Miho Furumichi, and Mao Tanabe. Kegg as a reference resource for gene and protein annotation.Nucleic acids research, 44(D1):D457–D462, 2016

    2016

  45. [53]

    Kriege, Franka Bause, Kristian Kersting, Petra Mutzel, and Marion Neumann

    Christopher Morris, Nils M. Kriege, Franka Bause, Kristian Kersting, Petra Mutzel, and Marion Neumann. Tudataset: A collection of benchmark datasets for learning with graphs. InICML 2020 Workshop on Graph Representation Learning and Beyond (GRL+ 2020), 2020. 11

    2020

Pith tools

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