pith. sign in

arxiv: 2605.21054 · v1 · pith:WYVCFT7Enew · submitted 2026-05-20 · 💻 cs.SI · cs.CY

DeTox-Fed: Detecting Toxic Conversations in the Fediverse with Federated Graph Neural Networks

Pantelitsa Leonidou , Nikos Salamanos , Sotiris Gypsiotis , Michael Sirivianos This is my paper

Pith reviewed 2026-05-21 01:53 UTC · model grok-4.3

classification 💻 cs.SI cs.CY
keywords federated learninggraph neural networkstoxic conversation detectionfediversedecentralized social networkscontent moderationpleromaprivacy preservation
0
0 comments X

The pith

A federated graph neural network detects toxic conversations in the Fediverse by training on local graphs of shared user participation without sharing raw text or labels.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents DeTox-Fed to tackle moderation in decentralized social networks where each instance holds only a partial view of conversations and follows its own policies. Each instance builds a local graph with conversation trees as nodes and shared users as edges, then trains a graph neural network in a federated manner that incorporates structure, interaction patterns, aggregate statistics, and sentiment signals. This setup produces a toxicity classifier that works across instances while keeping all raw conversations and labels local. A sympathetic reader would care because independent instances could gain reliable detection even when local labels are scarce and participation is incomplete.

Core claim

DeTox-Fed achieves stable toxic conversation detection under limited local labels, partial client participation, and varying moderation thresholds on a large Pleroma conversation dataset by training a Graph Neural Network in a federated setup on local conversation graphs built from shared user participation, combined with aggregate statistics and sentiment signals, without requiring instances to share raw conversations or moderation labels.

What carries the argument

The local conversation graph, with nodes as conversation trees and edges as shared user participation across instances, which supplies the structure and interaction signals that the federated GNN uses to classify toxicity.

If this is right

  • Toxic conversations can be identified reliably even when each instance possesses only limited local labels.
  • Detection remains stable when only a subset of instances participate in the federated training round.
  • The classifier tolerates different moderation thresholds chosen independently by each instance.
  • The approach supplies a concrete path toward semi-automated moderation that respects data locality in decentralized networks.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Smaller instances could obtain effective detection by borrowing patterns learned from larger ones without any direct data transfer.
  • The same graph-construction step might support detection of other conversation-level issues such as misinformation or harassment.
  • Deployment would benefit from live tests that measure how quickly the model adapts when new instances join the federation.
  • The framework naturally connects to other privacy-preserving machine-learning tasks on partially overlapping social graphs.

Load-bearing premise

Local conversation graphs built from shared user participation across instances, combined with aggregate statistics and sentiment, contain sufficient signal to classify toxicity without access to raw conversation text or cross-instance labels.

What would settle it

Running DeTox-Fed on the Pleroma dataset after removing all graph edges and sentiment aggregates, then checking whether detection performance falls to the level of a simple local text-only baseline.

Figures

Figures reproduced from arXiv: 2605.21054 by Michael Sirivianos, Nikos Salamanos, Pantelitsa Leonidou, Sotiris Gypsiotis.

Figure 1
Figure 1. Figure 1: Toy example of the DETOX-FED framework. (A) The raw data consist of conversation trees from different DSN instances, where each tree contains an initial post and its replies. Toots from the same user are shown with the same color. (B) For each DSN instance, we construct a weighted conversation graph in which nodes represent conversation trees and edge weights represent the number of users shared between tw… view at source ↗
Figure 2
Figure 2. Figure 2: LLM-based Toxicity Detection: The local setup [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Effect of noise-corrected backboning on conver [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
read the original abstract

The rise of decentralized social networks (DSNs), and in particular the rapid uptake of the Fediverse (e.g., Pleroma, Mastodon, Lemygrad), introduces new challenges in content moderation. Independent instances host their own data, follow different moderation policies, and often observe only partial views of conversations. We present DeTox-Fed, a federated graph-learning framework for detecting toxic conversations in DSNs without requiring instances to share raw conversations or moderation labels. Each instance constructs a local conversation graph, where nodes represent conversation trees and edges capture shared user participation across conversations. A Graph Neural Network (GNN) is then trained in a federated learning setup, allowing instances to collaboratively learn a toxicity classifier while preserving data locality. Unlike text-only moderation approaches, DeTox-Fed combines conversational structure, user-interaction patterns, conversation-level statistics, and aggregate sentiment signals. We evaluate the framework on a large Pleroma conversation dataset and show that it achieves stable toxic conversation detection under limited local labels, partial client participation, and varying moderation thresholds. Our results indicate that federated graph-based moderation is a promising direction for semi-automated moderation in decentralized social networks.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper introduces DeTox-Fed, a federated graph neural network framework for detecting toxic conversations in decentralized social networks such as the Fediverse (e.g., Pleroma). Each instance builds a local conversation graph with nodes as conversation trees and edges representing shared user participation across instances; these are augmented with aggregate statistics and sentiment signals. A GNN is trained collaboratively via federated learning without sharing raw conversation text or cross-instance labels. The central claim is that this yields stable toxic conversation detection on a large Pleroma dataset under limited local labels, partial client participation, and varying moderation thresholds.

Significance. If the empirical claims hold with verifiable metrics, the work would represent a meaningful step toward privacy-preserving, graph-based moderation tools for decentralized platforms where instances have only partial views and independent policies. It usefully combines conversational structure with federated learning and sentiment proxies, extending beyond purely text-based approaches. The absence of quantitative results, however, limits assessment of whether the graph component adds signal beyond simpler baselines.

major comments (2)
  1. [Evaluation] Evaluation section: the abstract and results description assert positive performance and stability under limited labels, partial participation, and varying thresholds, yet supply no quantitative metrics (precision, recall, F1, AUC), baselines, error bars, data-split details, or model-architecture specifications. This directly undermines verification of the central stability claim.
  2. [§4.3] §4.3 (or equivalent results/ablation subsection): the framework's headline result rests on the assumption that local conversation graphs plus aggregate sentiment suffice for toxicity classification. No ablation isolating the GNN/graph contribution from the sentiment component is reported; if performance is driven primarily by sentiment, the federated graph-learning framing does not support the claimed advance on the Pleroma dataset.
minor comments (2)
  1. [Abstract] Abstract: key numerical results (e.g., F1 scores or stability ranges) should be included to allow readers to gauge the strength of the reported stability.
  2. [§3] Notation: the precise definition of 'conversation tree' nodes and 'shared user participation' edges should be formalized with a small diagram or pseudocode to clarify how the local graph is constructed from instance data.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive feedback on our manuscript. The comments highlight important gaps in the presentation of empirical results, and we will revise the paper to address them directly.

read point-by-point responses

  1. Referee: [Evaluation] Evaluation section: the abstract and results description assert positive performance and stability under limited labels, partial participation, and varying thresholds, yet supply no quantitative metrics (precision, recall, F1, AUC), baselines, error bars, data-split details, or model-architecture specifications. This directly undermines verification of the central stability claim.

    Authors: We agree that the current version of the manuscript does not present the requested quantitative metrics, baselines, error bars, data-split details, or full architecture specifications in the evaluation section. In the revised manuscript we will add a dedicated results subsection containing precision, recall, F1, and AUC values for the federated GNN under the stated conditions (limited labels, partial participation, varying thresholds), together with comparisons against non-graph and non-federated baselines, standard-error bars from repeated runs, explicit train/validation/test splits, and the precise GNN layer and aggregation choices. revision: yes

  2. Referee: [§4.3] §4.3 (or equivalent results/ablation subsection): the framework's headline result rests on the assumption that local conversation graphs plus aggregate sentiment suffice for toxicity classification. No ablation isolating the GNN/graph contribution from the sentiment component is reported; if performance is driven primarily by sentiment, the federated graph-learning framing does not support the claimed advance on the Pleroma dataset.

    Authors: We acknowledge that an ablation isolating the graph and GNN contribution from the sentiment signals is necessary to substantiate the claimed advance. In the revised version we will include an ablation study that reports performance when the graph edges and GNN are removed (reducing to a federated sentiment-only classifier) and when sentiment features are removed (reducing to a graph-only model). These comparisons will be presented alongside the full DeTox-Fed results on the Pleroma dataset to demonstrate the incremental value of the conversational graph structure. revision: yes

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper introduces DeTox-Fed as a new federated graph-learning framework that constructs local conversation graphs from shared user participation, combines them with aggregate statistics and sentiment, and trains a GNN under federated learning to detect toxicity. Evaluation occurs on an external large Pleroma dataset with reported performance under limited labels and partial participation. No mathematical equations, derivations, or first-principles results are claimed that reduce performance metrics to fitted parameters or self-definitions from the same inputs. Central claims rest on empirical results from the described construction rather than any self-citation chain, ansatz smuggling, or renaming of known patterns. The framework is presented as an independent construction evaluated externally.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption that user-participation graphs plus aggregate signals suffice for toxicity classification when raw text and labels cannot be shared; no free parameters or invented entities are explicitly introduced in the abstract.

axioms (1)
  • domain assumption Conversation graphs constructed from shared user participation across instances capture toxicity-relevant structure
    Invoked when the framework defines nodes as conversation trees and edges as shared user participation.

pith-pipeline@v0.9.0 · 5754 in / 1239 out tokens · 30914 ms · 2026-05-21T01:53:48.067721+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

108 extracted references · 108 canonical work pages

  1. [1]

    Artificial intelligence and statistics , pages=

    Communication-efficient learning of deep networks from decentralized data , author=. Artificial intelligence and statistics , pages=. 2017 , organization=

    work page 2017

  2. [2]

    Proceedings of the international AAAI conference on web and social media , volume=

    HyperGraphDis: Leveraging hypergraphs for contextual and social-based disinformation detection , author=. Proceedings of the international AAAI conference on web and social media , volume=

    work page

  3. [3]

    International Conference on Data Engineering (ICDE) , year =

    Michele Coscia and Frank Neffke , title =. International Conference on Data Engineering (ICDE) , year =

    work page

  4. [4]

    Proceedings of the International AAAI Conference on Web and Social Media , volume=

    Decentralised Moderation for Interoperable Social Networks: A Conversation-based Approach for Pleroma and the Fediverse , author=. Proceedings of the International AAAI Conference on Web and Social Media , volume=

    work page

  5. [5]

    Bin Zia, Haris and Raman, Aravindh and Castro, Ignacio and Hassan Anaobi, Ishaku and De Cristofaro, Emiliano and Sastry, Nishanth and Tyson, Gareth , title =. Proc. ACM Meas. Anal. Comput. Syst. , month = jun, articleno =. 2022 , issue_date =. doi:10.1145/3530901 , abstract =

    work page doi:10.1145/3530901 2022

  6. [6]

    and Sastry, Nishanth , title =

    Agarwal, Vibhor and Joglekar, Sagar and Young, Anthony P. and Sastry, Nishanth , title =. Proceedings of the ACM Web Conference 2022 , pages =. 2022 , isbn =. doi:10.1145/3485447.3512144 , abstract =

    work page doi:10.1145/3485447.3512144 2022

  7. [7]

    arXiv preprint arXiv:2501.05871 , year=

    Collaborative Content Moderation in the Fediverse , author=. arXiv preprint arXiv:2501.05871 , year=

    work page arXiv

  8. [8]

    Journal of Complex Networks , volume=

    Bayesian inference of network structure from unreliable data , author=. Journal of Complex Networks , volume=. 2020 , publisher=

    work page 2020

  9. [9]

    Federated Graph Neural Networks: Overview, Techniques, and Challenges , year=

    Liu, Rui and Xing, Pengwei and Deng, Zichao and Li, Anran and Guan, Cuntai and Yu, Han , journal=. Federated Graph Neural Networks: Overview, Techniques, and Challenges , year=

    work page

  10. [10]

    Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining , pages=

    Deepwalk: Online learning of social representations , author=. Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining , pages=

    work page

  11. [11]

    XLM - T : Multilingual Language Models in T witter for Sentiment Analysis and Beyond

    Barbieri, Francesco and Espinosa Anke, Luis and Camacho-Collados, Jose. XLM - T : Multilingual Language Models in T witter for Sentiment Analysis and Beyond. Proceedings of the Thirteenth Language Resources and Evaluation Conference. 2022

    work page 2022

  12. [12]

    NeurIPS Workshop on Scalability, Privacy, and Security in Federated Learning , year=

    FedML: A Research Library and Benchmark for Federated Machine Learning , author=. NeurIPS Workshop on Scalability, Privacy, and Security in Federated Learning , year=

    work page

  13. [13]

    2021 , eprint=

    FedGraphNN: A Federated Learning System and Benchmark for Graph Neural Networks , author=. 2021 , eprint=

    work page 2021

  14. [14]

    Measuring Online Social Bubbles , volume =

    Nikolov, Dimitar and Oliveira, Diego and Flammini, Alessandro and Menczer, Filippo , year =. Measuring Online Social Bubbles , volume =. PeerJ Computer Science , doi =

    work page

  15. [15]

    and Van Der Heide, Brandon , title =

    Westerman, David and Spence, Patric R. and Van Der Heide, Brandon , title =. 2014 , issue_date =. doi:10.1111/jcc4.12041 , journal =

    work page doi:10.1111/jcc4.12041 2014

  16. [16]

    Reliance on Facebook for news and its influence on political engagement , volume =

    David, Clarissa and San Pascual, Ma and Torres, Ma , year =. Reliance on Facebook for news and its influence on political engagement , volume =. PLOS ONE , doi =

    work page

  17. [17]

    Social impacts in social media: An examination of perceived truthfulness and sharing of information , volume =

    Li, Huaye and Sakamoto, Yasuaki , year =. Social impacts in social media: An examination of perceived truthfulness and sharing of information , volume =. Computers in Human Behavior , doi =

    work page

  18. [18]

    Social Clicks: What and Who Gets Read on Twitter? , doi =

    Gabielkov, Maksym and Ramachandran, Arthi and Chaintreau, Augustin and Legout, Arnaud , year =. Social Clicks: What and Who Gets Read on Twitter? , doi =

    work page

  19. [19]

    African Identities , volume =

    Vincent Chenzi , title =. African Identities , volume =. 2021 , publisher =. doi:10.1080/14725843.2020.1804321 , URL =

    work page doi:10.1080/14725843.2020.1804321 2021

  20. [20]

    Political and Social Impact of digital Fake news in an era of Social Media , isbn =

    Safieddine, Fadi , year =. Political and Social Impact of digital Fake news in an era of Social Media , isbn =

    work page

  21. [21]

    Fake News, Real Problems for Brands: The Impact of Content Truthfulness and Source Credibility on consumers' Behavioral Intentions toward the Advertised Brands , volume =

    Visentin, Marco and Pizzi, Gabriele and Pichierri, Marco , year =. Fake News, Real Problems for Brands: The Impact of Content Truthfulness and Source Credibility on consumers' Behavioral Intentions toward the Advertised Brands , volume =

    work page

  22. [22]

    The Influence of Presumed Fake News Influence: Examining Public Support for Corporate Corrective Response, Media Literacy Interventions, and Governmental Regulation , journal =

    Cheng, Yang and Chen, Zifei Fay , year =. The Influence of Presumed Fake News Influence: Examining Public Support for Corporate Corrective Response, Media Literacy Interventions, and Governmental Regulation , journal =

    work page

  23. [23]

    Fake News as a Threat to National Security , volume =

    Belova, Gabriela and Georgieva, Gergana , year =. Fake News as a Threat to National Security , volume =. International conference KNOWLEDGE-BASED ORGANIZATION , doi =

    work page

  24. [24]

    A Comprehensive Survey on Graph Neural Networks , volume =

    Wu, Zonghan and Pan, Shirui and Chen, Fengwen and Long, Guodong and Zhang, Chengqi and Yu, Philip , year =. A Comprehensive Survey on Graph Neural Networks , volume =. IEEE Transactions on Neural Networks and Learning Systems , doi =

    work page

  25. [25]

    2020 , author =

    AI Open , volume =. 2020 , author =

    work page 2020

  26. [26]

    Graph convolutional networks: a comprehensive review , volume =

    Zhang, Si and Tong, Hanghang and Xu, Jiejun and Maciejewski, Ross , year =. Graph convolutional networks: a comprehensive review , volume =

    work page

  27. [27]

    FakeBERT: Fake news detection in social media with a BERT-based deep learning approach , journal =

    Kaliyar, Rohit and Goswami, Anurag and Narang, Pratik , year =. FakeBERT: Fake news detection in social media with a BERT-based deep learning approach , journal =

    work page

  28. [28]

    Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies,

    Devlin, Jacob and Chang, Ming-Wei and Lee, Kenton and Toutanova, Kristina , year =. Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies,

    work page 2019

  29. [29]

    Detection of Online Fake News Using N-Gram Analysis and Machine Learning Techniques

    Ahmed, Hadeer and Traore, Issa and Saad, Sherif. Detection of Online Fake News Using N-Gram Analysis and Machine Learning Techniques. Intelligent, Secure, and Dependable Systems in Distributed and Cloud Environments. 2017

    work page 2017

  30. [30]

    Towards automatic fake news classification , volume =

    Ghosh, Souvick and Shah, Chirag , year =. Towards automatic fake news classification , volume =

    work page

  31. [31]

    Advances in Neural Information Processing Systems , title =

    Zhang, Xiang and Zhao, Junbo and Lecun, Yann , year =. Advances in Neural Information Processing Systems , title =

    work page

  32. [32]

    A Survey of Fake News: Fundamental Theories, Detection Methods, and Opportunities , volume =

    Zhou, Xinyi and Zafarani, Reza , year =. A Survey of Fake News: Fundamental Theories, Detection Methods, and Opportunities , volume =. ACM Computing Surveys , doi =

    work page

  33. [33]

    Mark my words!: linguistic style accommodation in social media , url =

    Castillo, Carlos and Mendoza, Marcelo and Poblete, Barbara , title =. Proceedings of the 20th International Conference on World Wide Web , pages =. 2011 , isbn =. doi:10.1145/1963405.1963500 , abstract =

    work page doi:10.1145/1963405.1963500 2011

  34. [34]

    Proceedings of the 2019 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining , pages =

    Shu, Kai and Zhou, Xinyi and Wang, Suhang and Zafarani, Reza and Liu, Huan , title =. Proceedings of the 2019 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining , pages =. 2020 , isbn =. doi:10.1145/3341161.3342927 , abstract =

    work page doi:10.1145/3341161.3342927 2019

  35. [35]

    2022 , eprint=

    Combating Health Misinformation in Social Media: Characterization, Detection, Intervention, and Open Issues , author=. 2022 , eprint=

    work page 2022

  36. [36]

    2023 , eprint=

    A Golden Age: Conspiracy Theories' Relationship with Misinformation Outlets, News Media, and the Wider Internet , author=. 2023 , eprint=

    work page 2023

  37. [37]

    2023 , eprint=

    Sub-Standards and Mal-Practices: Misinformation's Role in Insular, Polarized, and Toxic Interactions , author=. 2023 , eprint=

    work page 2023

  38. [38]

    Swinging in the States: Does disinformation on Twitter mirror the US presidential election system? , doi =

    Pratelli, Manuel and Petrocchi, Marinella and Saracco, Fabio and De Nicola, Rocco , year =. Swinging in the States: Does disinformation on Twitter mirror the US presidential election system? , doi =

    work page

  39. [39]

    2023 , eprint=

    A Data-driven Understanding of Left-Wing Extremists on Social Media , author=. 2023 , eprint=

    work page 2023

  40. [40]

    Misleading Repurposing on Twitter , volume =

    Elmas, Tuğrulcan and Overdorf, Rebekah and Aberer, Karl , year =. Misleading Repurposing on Twitter , volume =. Proceedings of the International AAAI Conference on Web and Social Media , doi =

    work page

  41. [41]

    A Unified Graph-Based Approach to Disinformation Detection Using Contextual and Semantic Relations , volume =

    Paraschiv, Marius and Salamanos, Nikos and Iordanou, Costas and Laoutaris, Nikolaos and Sirivianos, Michael , year =. A Unified Graph-Based Approach to Disinformation Detection Using Contextual and Semantic Relations , volume =. Proceedings of the International AAAI Conference on Web and Social Media , doi =

    work page

  42. [42]

    2017 , eprint=

    Some Like it Hoax: Automated Fake News Detection in Social Networks , author=. 2017 , eprint=

    work page 2017

  43. [43]

    Proceedings of the 27th ACM International Conference on Information and Knowledge Management , pages =

    Guo, Han and Cao, Juan and Zhang, Yazi and Guo, Junbo and Li, Jintao , title =. Proceedings of the 27th ACM International Conference on Information and Knowledge Management , pages =. 2018 , isbn =. doi:10.1145/3269206.3271709 , abstract =

    work page doi:10.1145/3269206.3271709 2018

  44. [44]

    Multimodal Fusion with Recurrent Neural Networks for Rumor Detection on Microblogs , doi =

    Jin, Zhiwei and Cao, Juan and Guo, Han and Zhang, Yongdong and Luo, Jiebo , year =. Multimodal Fusion with Recurrent Neural Networks for Rumor Detection on Microblogs , doi =

    work page

  45. [45]

    Evaluating Event Credibility on Twitter , journal =

    Gupta, Manish and Zhao, Peixiang and Han, Jiawei , year =. Evaluating Event Credibility on Twitter , journal =

    work page

  46. [46]

    Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining , pages =

    Shu, Kai and Cui, Limeng and Wang, Suhang and Lee, Dongwon and Liu, Huan , title =. Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining , pages =. 2019 , isbn =. doi:10.1145/3292500.3330935 , abstract =

    work page doi:10.1145/3292500.3330935 2019

  47. [47]

    Proceedings of the Eleventh ACM International Conference on Web Search and Data Mining , pages =

    Wu, Liang and Liu, Huan , title =. Proceedings of the Eleventh ACM International Conference on Web Search and Data Mining , pages =. 2018 , isbn =. doi:10.1145/3159652.3159677 , abstract =

    work page doi:10.1145/3159652.3159677 2018

  48. [48]

    Proceedings of the Twelfth ACM International Conference on Web Search and Data Mining , pages =

    Shu, Kai and Wang, Suhang and Liu, Huan , title =. Proceedings of the Twelfth ACM International Conference on Web Search and Data Mining , pages =. 2019 , isbn =. doi:10.1145/3289600.3290994 , abstract =

    work page doi:10.1145/3289600.3290994 2019

  49. [49]

    Proceedings of the International AAAI Conference on Web and Social Media , author=

    Hierarchical Propagation Networks for Fake News Detection: Investigation and Exploitation , volume=. Proceedings of the International AAAI Conference on Web and Social Media , author=. 2020 , month=. doi:10.1609/icwsm.v14i1.7329 , abstractNote=

    work page doi:10.1609/icwsm.v14i1.7329 2020

  50. [50]

    Proceedings of the 29th ACM International Conference on Information & Knowledge Management , pages =

    Nguyen, Van-Hoang and Sugiyama, Kazunari and Nakov, Preslav and Kan, Min-Yen , title =. Proceedings of the 29th ACM International Conference on Information & Knowledge Management , pages =. 2020 , publisher =

    work page 2020

  51. [51]

    2020 , howpublished =

    MBFC , title =. 2020 , howpublished =

    work page 2020

  52. [52]

    2020 , howpublished =

    PolitiFacts , title =. 2020 , howpublished =

    work page 2020

  53. [53]

    2020 , howpublished =

    FactCheck , title =. 2020 , howpublished =

    work page 2020

  54. [54]

    2020 , howpublished =

    Snopes , title =. 2020 , howpublished =

    work page 2020

  55. [55]

    2021 , howpublished =

    Bitly , title =. 2021 , howpublished =

    work page 2021

  56. [56]

    2021 , howpublished =

    TinyURL , title =. 2021 , howpublished =

    work page 2021

  57. [57]

    2016 , howpublished =

    Mark Davis , title =. 2016 , howpublished =

    work page 2016

  58. [58]

    2020 , howpublished =

    Maciej Szpakowski , title =. 2020 , howpublished =

    work page 2020

  59. [59]

    2021 , howpublished =

    Peter Macinec , title =. 2021 , howpublished =

    work page 2021

  60. [60]

    Graph similarity scoring and matching , volume =

    Zager, Laura and Verghese, George , year =. Graph similarity scoring and matching , volume =

    work page

  61. [61]

    A Measure of Similarity between Graph Vertices: Applications to Synonym Extraction and Web Searching , volume =

    Blondel, Vincent and Gajardo, Anahi and Heymans, Maureen and Senellart, Pierre and Van Dooren, Paul , year =. A Measure of Similarity between Graph Vertices: Applications to Synonym Extraction and Web Searching , volume =. SIAM Review , doi =

    work page

  62. [62]

    A Novel Method for Graph Matching , doi =

    Zhao, Chunyu and Yin, Zhaoning and Wang, Haoru and Zhao, Xiuyang and Niu, Dongmei , year =. A Novel Method for Graph Matching , doi =

    work page

  63. [63]

    Graph matching , isbn =

    Bisseling, Rob , year =. Graph matching , isbn =

    work page

  64. [64]

    Gomaa and Aly A

    Wael H. Gomaa and Aly A. Fahmy , title =. International Journal of Computer Applications , year =

    work page

  65. [65]

    International Conference on Learning Representations , year=

    Graph Neural Networks Exponentially Lose Expressive Power for Node Classification , author=. International Conference on Learning Representations , year=

    work page

  66. [66]

    Deeper Insights into Graph Convolutional Networks for Semi-Supervised Learning , url=

    Li, Qimai and Han, Zhichao and Wu, Xiao-Ming , year =. Deeper Insights into Graph Convolutional Networks for Semi-Supervised Learning , url=

    work page

  67. [67]

    Revisiting Graph Neural Networks: All We Have is Low-Pass Filters , url=

    NT, Hoang and Maehara, Takanori , year =. Revisiting Graph Neural Networks: All We Have is Low-Pass Filters , url=

    work page

  68. [68]

    On the Bottleneck of Graph Neural Networks and its Practical Implications , url=

    Alon, Uri and Yahav, Eran , year =. On the Bottleneck of Graph Neural Networks and its Practical Implications , url=

    work page

  69. [69]

    Proceedings of the 29th ACM International Conference on Information and Knowledge Management (CIKM '20) , organization =

    Benedek Rozemberczki and Oliver Kiss and Rik Sarkar , year =. Proceedings of the 29th ACM International Conference on Information and Knowledge Management (CIKM '20) , organization =

    work page

  70. [70]

    Proceedings of the 20th International Conference on Knowledge Discovery and Data Mining , pages =

    Perozzi, Bryan and Al-Rfou, Rami and Skiena, Steven , title =. Proceedings of the 20th International Conference on Knowledge Discovery and Data Mining , pages =. 2014 , publisher =

    work page 2014

  71. [71]

    powerlaw: A Python Package for Analysis of Heavy-Tailed Distributions , year =

    Alstott, Jeff AND Bullmore, Ed AND Plenz, Dietmar , journal =. powerlaw: A Python Package for Analysis of Heavy-Tailed Distributions , year =

    work page

  72. [72]

    Watts , title =

    Sharad Goel and Ashton Anderson and Jake Hofman and Duncan J. Watts , title =. Management Science , volume =. 2015 , number =

    work page 2015

  73. [73]

    , year =

    Veličković, Petar and Cucurull, Guillem and Casanova, Arantxa and Romero, Adriana and Lio, Pietro and Bengio, Y. , year =

    work page

  74. [74]

    Hypergraph convolution and hypergraph attention , volume =

    Bai, Song and Zhang, Feihu and Torr, Philip , year =. Hypergraph convolution and hypergraph attention , volume =. Pattern Recognition , doi =

    work page

  75. [75]

    and Welling, Max , biburl =

    Kipf, Thomas N. and Welling, Max , biburl =. Proceedings of the 5th International Conference on Learning Representations , interhash =

    work page

  76. [76]

    Inductive Representation Learning on Large Graphs , volume =

    Hamilton, Will and Ying, Zhitao and Leskovec, Jure , booktitle =. Inductive Representation Learning on Large Graphs , volume =

    work page

  77. [77]

    , booktitle=

    Fey, Matthias and Lenssen, Jan E. , booktitle=. Fast Graph Representation Learning with

    work page

  78. [78]

    2021 , publisher =

    bab2min and Douglas Fenstermacher , title =. 2021 , publisher =

    work page 2021

  79. [79]

    and Jordan, Michael I

    Blei, David M. and Jordan, Michael I. and Griffiths, Thomas L. and Tenenbaum, Joshua B. , title =. 2003 , publisher =

    work page 2003

  80. [80]

    Teh, Y. W. and Jordan, M. I. and Beal, M. J. and Blei D. M. , title =. 2005 , publisher =

    work page 2005

Showing first 80 references.