Pith. sign in

REVIEW 5 major objections 5 minor 3 cited by

Rethinking Graph Contrastive Learning through Relative Similarity Preservation

T0 review · 5 major / 5 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 The paper establishes that label consistency in graphs decays systematically with structural distance, and builds a contrastive learning method that preserves this relative similarity instead of enforcing similarity between artificial…

desk verdict The empirical decay pattern and RELGCL losses are worth a look, but the random-walk theory is presented as proving something it does not prove, and the paper never connects the theoretical quantity to the measured one. read the letter →

arxiv 2505.05533 v2 pith:NCG4DNKE submitted 2025-05-08 cs.LG cs.AI

Zhiyuan Ning , Pengfei Wang , Ziyue Qiao , Pengyang Wang , Yuanchun Zhou This is my paper · ORCID
classification cs.LGcs.AI
keywords graphcontrastivelearningrelativesimilaritylabelconsistencydecayrandomwalktheoryhomophilyandheterophilycollectiveobjectivenodeclassificationself-supervised
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

Across 11 real-world graphs, the paper finds a universal statistical pattern: the fraction of k-hop neighbors sharing a node's label shrinks as k grows, decaying smoothly in homophily graphs and with oscillations in heterophily graphs. It argues this ordered decay is the natural similarity signal graphs encode, and that graph contrastive learning (GCL) should preserve it rather than force absolute similarity between augmented views. To ground the pattern, the paper proves via random walk theory that the same-label return probability of a degree-weighted walk converges exponentially to the edge-proportion of the anchor's label. On this basis it builds RELGCL, whose pairwise and listwise objectives keep near-hop neighbors collectively more similar to the anchor than farther hops. If the pattern holds, structural proximity can replace view augmentation as the organizing principle for GCL.

What carries the argument

The carrying object is label consistency as a function of hop distance. Empirically it is LC_emp(k) from Eq. (1), the graph-averaged fraction of exact k-hop neighbors sharing the anchor's label; theoretically it is LC_prob(k) = p_k(i|i), the same-label return probability of a degree-weighted random walk with self-loops, aggregated to a label-level transition matrix T. The load-bearing identities are Theorem 1, which gives exponential convergence of label probabilities to the stationary edge-proportion distribution π, and Corollary 1, which turns that convergence into a decay statement for LC_prob(k); Proposition 1 uses the sign of the second eigenvalue of T in a two-label model to explain smooth versus oscillatory decay. The method side is the collective similarity ratio: summing exponentials inside the logarithm (Lin) so optimization acts on aggregate similarity of a group of near neighbors versus a group of farther neighbors, capped by threshold α, with RELGCL-PAIR comparing individual hop pairs and RELGCL-LIST comparing each hop against all later hops.

What would settle it

Compute LC_emp(k) from Eq. (1) on a connected even cycle whose labels alternate A,B,A,B,...: hop-1 neighbors are all different-label (LC=0), but hop-2 neighbors are all same-label (LC=1), so the averaged label consistency rises rather than falls with distance; that graph would refute the claimed universal decay if it appears as a counterexample or is reproducible synthetically. A milder test: any real graph where the graph-averaged LC_emp sequence over k=1..5 is not non-increasing overall.

Watch

Extended reading notes

Core claim

The paper's central claim is that label consistency — the average proportion of same-labeled nodes at each structural distance — diminishes systematically with distance in both homophily and heterophily graphs, and that this decay is a theorem-level consequence of random walk mixing. The theoretical object is LC_prob(k), the probability that a random walk with self-loops returns to the anchor's label after k steps: with one self-loop per node and a connected graph, the label-level transition matrix is irreducible and aperiodic, so LC_prob(k) converges exponentially to π_i, the degree-weighted share of label i, and π_i < 1 whenever other labels exist. The sign of the second eigenvalue of the label transition matrix then distinguishes smooth decay (homophily, λ2 > 0) from oscillatory decay (heterophily, λ2 < 0). RELGCL translates this statistical ordering into representation space by maximizing collective similarity ratios between nearer and farther hop-neighbor groups, in pairwise and listwise forms, and the paper reports that both outperform 20 baselines on homophily, heterophily, and large-scale graphs. The intended conclusion is that relative similarity, not absolute augmented-view similarity, is the signal GCL should learn.

Load-bearing premise

The load-bearing premise is that the empirical quantity LC_emp(k), computed from exact shortest-path hop neighborhoods, behaves like the theoretical quantity LC_prob(k), computed from a degree-weighted random walk with self-loops; the paper does not show the two measures coincide or decay identically.

Editorial extensions

If this is right

  • If the claim is right, graph contrastive learning does not need augmentation views: preserving ordered collective similarity across hops is a sufficient self-supervised objective.
  • The random walk guarantee implies that in any connected graph with self-loops and at least two labels, same-label return probability eventually tracks the degree-weighted label share, so relative similarity is a universal structural signal rather than a homophily artifact.
  • The eigenvalue explanation predicts heterophily graphs need a wider hop range for their oscillatory pattern to be modeled, which the paper's k-sensitivity results support.
  • Collective (sum-inside-log) objectives should beat individual positive-pair objectives whenever the semantic relation is statistical, as the paper's ablation on Lout versus Lin shows.

Reading between the lines

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

  • The same decay law could be tested as a theory of oversmoothing in GNNs: layer-wise message passing approximates random-walk mixing, so representation similarity should converge to label edge-proportions, predicting when deeper models lose discriminability.
  • The two objectives are ranking losses over hop groups; they should transfer to link prediction or graph-level classification, but the paper only evaluates node-level tasks, so that extension is untested.
  • Spectral computation of the label transition matrix's second eigenvalue on any given graph would let practitioners predict smooth versus oscillatory decay and choose between the pairwise and listwise variants before training.
  • If LC_prob and LC_emp diverge on specific graphs, the method's motivation would still stand on the empirical pattern, but the theorem would need replacement by a concentration argument over hop neighborhoods.
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

5 major / 5 minor

Summary. The paper studies label consistency in graphs, defined as the average fraction of k-hop neighbors sharing the anchor's label. It claims to discover a universal pattern: label consistency systematically decays with structural distance in both homophily and heterophily graphs, manifesting as smooth decay in the former and oscillatory decay in the latter. The authors propose a random-walk convergence theorem as a theoretical guarantee for this pattern, and then introduce RELGCL, a graph contrastive learning framework that preserves the relative similarity between different hop neighborhoods via pairwise and listwise collective-similarity objectives. Experiments on 11 datasets compare against 20 baselines and report consistent improvements.

Significance. If the central claims were correct, the paper would offer a principled alternative to the absolute-similarity paradigm in graph contrastive learning, replacing augmented views with structural-proximity-based relative similarity. The empirical study is broad, the method is clearly described, and the comparison with 20 baselines is extensive. However, the theoretical guarantees do not apply to the empirical quantity they are claimed to explain, and the method strongly resembles the authors' own earlier neighborhood-ranking work. As it stands, the paper's main novelty—the claimed theoretical foundation—is not established, and the incremental contribution over prior work is not clearly delineated.

major comments (5)
  1. [§2.2 and Appendix A.2] The random-walk theory governs LC_prob(k), a degree-weighted label-level return probability with self-loops, whereas the empirical measure LC_emp(k) in Eq. (1) is a uniform average over anchors of the fraction of exact shortest-path k-hop neighbors sharing the anchor's label. No argument in Section 2.2 or Appendix A identifies these two quantities: walks may revisit nodes, are not restricted to shortest paths, and are degree-weighted, and the self-loop assumption in Lemma 1 has no counterpart in Eq. (1). Therefore Theorem 1 and Corollary 1 do not establish the claimed decay of LC_emp.
  2. [§2.2, Corollary 1 and Proposition 1] Even for LC_prob, the stated result is only exponential convergence to π_i < 1; convergence does not imply monotone or overall decay. In the paper's own two-label heterophily model T ≈ [[1−p, p], [p, 1−p]] with p > 0.5, LC_prob(1) = 1−p < 0.5 while π_i = 0.5 for balanced classes, so the quantity increases toward its limit. The 'oscillatory decay' claim is thus not a consequence of the theorem, and the proof of Corollary 1 only bounds |LC_prob(k) − π_i|.
  3. [§3.1, Eq. (7)] The inequality E[simstat(v,n)] > E[simstat(v,>n)] is asserted as a proven consequence of Section 2, but it is neither proven by the random-walk theorems (which concern a different quantity and only convergence) nor derived from the definitions in Eqs. (5)-(6). The RELGCL objectives are then built on this unproven inequality, so the claimed principled foundation for the method is not established; Eq. (7) is an additional empirical assumption.
  4. [Appendix A.6] The appendix states that 'a complete theoretical characterization accounting for multiple labels, arbitrary graph structures, and varying degrees of homophily/heterophily remains an open challenge.' This limitation statement directly contradicts the abstract's claim that the paper establishes 'theoretical guarantees' for the universal decay pattern. The manuscript is internally inconsistent about the strength of its theoretical contribution.
  5. [§3.3 and Related Work] The proposed RELGCL objectives—comparing the collective similarity of one hop against further hops using ratio-based InfoNCE-style losses—closely resemble the authors' own 'Graph soft-contrastive learning via neighborhood ranking' [Ning et al., 2022], which is cited only in passing and is not included as a baseline. The paper does not discuss the differences from this prior work, so the claimed novelty and the conclusion that RELGCL outperforms '20 existing approaches' are overstated without a direct comparison against the most relevant predecessor.
minor comments (5)
  1. [§2.1, Eq. (1)] The notation N(vi)[k] is defined as a set of sets {N(vi)[1], ..., N(vi)[k], N(vi)[k+1]}, but in Eq. (1) it is used as a set of nodes; please clarify the intended meaning and use distinct notation for the union of the first k hops.
  2. [Table 7] There is a typo in the table header: 'Homophiliy' should be 'Homophily'.
  3. [Appendix C.4] The paper does not report the per-dataset hyperparameter values (k, α, τ, τ spacing, embedding dimension) that were selected by grid search; providing a supplementary table would substantially improve reproducibility.
  4. [Figure 2] The figure shows point estimates without error bars or standard deviations; given that the 'oscillatory decay' for heterophily graphs rests on small differences between hops, confidence intervals would help assess the robustness of the pattern.
  5. [Tables 1-3] Several improvements over the best baseline are within one standard deviation (e.g., Chameleon in Table 2: 69.25±0.89 vs 69.12±0.24); reporting confidence intervals or statistical significance tests would strengthen the claim of consistent superiority.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the decay pattern is measured from labels, the theory is standard Markov chain convergence, and RELGCL's loss is built from hop memberships without labels or fitted constants.

full rationale

The paper's central derivation is not circular. The universal decay pattern is defined empirically via Eq. (1) (LC_emp) and measured on 11 datasets; it is not generated by the method. The theoretical support (Theorem 1, Corollary 1) is a standard irreducible-aperiodic Markov chain convergence result (Perron-Frobenius, cited to Levin and Peres 2017), not an assumption equivalent to the conclusion. The RELGCL objectives (Eqs. 11-14) are constructed from k-hop neighborhood memberships and embedding similarities only; they do not use label consistency values or fitted constants, so the reported classification improvements are independent empirical evidence. The paper does contain a genuine gap that is not circularity: LC_emp counts exact shortest-path hop neighborhoods while LC_prob tracks degree-weighted walk-return probabilities with self-loops, and no proof identifies the two; Corollary 1's convergence bound does not by itself establish monotone or overall decay for all k, and the paper's own heterophily model (Proposition 1) has LC_prob(1) below its limit, i.e., upward convergence. The paper also admits in Appendix A.6 that 'a complete theoretical characterization remains an open challenge' and that real-world graphs may deviate from the idealized two-label assumptions. Self-citations (e.g., Ning et al. 2022, a neighborhood-ranking predecessor) appear in related work and are not load-bearing; no uniqueness theorem from the authors is invoked. No equation in the paper reduces to its own input, and no fitted parameter is renamed as a prediction, so the circularity score is low.

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

The theoretical claim rests on standard Markov chain theory plus an unstated identification between the empirical label consistency measure (Eq. 1) and the random-walk label return probability LC_prob. That identification is the weakest point: the former is an unweighted average over exact shortest-path k-hop neighborhoods, while the latter is a degree-weighted random walk probability with self-loops. The method itself adds tuned hyperparameters k, alpha, tau, and per-hop spacing rather than fitted physical constants.

free parameters (4)
  • neighborhood range k = 4 (grid-searched over 1-4)
    Sets how many hops are treated as ordered positives in the loss; selected per dataset by grid search on validation, so performance comparisons depend on this choice.
  • similarity threshold alpha = per-dataset optimum in [0.0001, 1.0]
    Clips the similarity ratio before taking the log; it changes the strength of relative similarity constraints and must be tuned per dataset.
  • temperature tau = grid 0.1-0.9
    Standard contrastive temperature in the cosine similarity exponentials; tuned per dataset.
  • per-hop temperature spacing = from {0, 0.0125, 0.025, 0.05, 0.1}
    Optional scaling between hop levels in the exponential terms; chosen by grid search.
assumptions (5)
  • standard math Perron-Frobenius theorem and Markov chain convergence theorem
    Invoked in Appendix A.4 to prove Theorem 1; these are standard and correct as stated.
  • domain assumption The graph is connected and each node has a self-loop
    Used in Lemma 1 to guarantee irreducibility and aperiodicity; real datasets lack self-loops, so the theoretical walker and the empirical hop counts diverge.
  • ad hoc to paper LC_emp(k) is the same object as LC_prob(k), the k-step random walk return probability aggregated by degree
    Section 2.2 introduces LC_prob to explain Eq. 1, but no derivation connects unweighted exact-hop proportions to degree-weighted walk probabilities with self-loops and revisits.
  • ad hoc to paper Two-label symmetric transition model approximates real homophily and heterophily graphs
    Appendix A.6 uses label symmetry and uniform transition assumptions, then states the full characterization remains open; the main text still presents the dichotomy as a general mechanism.
  • ad hoc to paper Standard Markov chain convergence implies monotone decay of label consistency
    Corollary 1 asserts decay, but the proof only establishes exponential convergence to a stationary value, which permits oscillation and overshoot.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Rethinking Graph Contrastive Learning through Relative Similarity Preservation." pith.science (2026). https://pith.science/paper/NCG4DNKE

@misc{pith2026250505533,
  author       = {Pith},
  title        = {Pith review of: Rethinking Graph Contrastive Learning through Relative Similarity Preservation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NCG4DNKE}},
  note         = {Machine review of arXiv:2505.05533}
}
read the original abstract

Graph contrastive learning (GCL) has achieved remarkable success by following the computer vision paradigm of preserving absolute similarity between augmented views. However, this approach faces fundamental challenges in graphs due to their discrete, non-Euclidean nature -- view generation often breaks semantic validity and similarity verification becomes unreliable. Through analyzing 11 real-world graphs, we discover a universal pattern transcending the homophily-heterophily dichotomy: label consistency systematically diminishes as structural distance increases, manifesting as smooth decay in homophily graphs and oscillatory decay in heterophily graphs. We establish theoretical guarantees for this pattern through random walk theory, proving label distribution convergence and characterizing the mechanisms behind different decay behaviors. This discovery reveals that graphs naturally encode relative similarity patterns, where structurally closer nodes exhibit collectively stronger semantic relationships. Leveraging this insight, we propose RELGCL, a novel GCL framework with complementary pairwise and listwise implementations that preserve these inherent patterns through collective similarity objectives. Extensive experiments demonstrate that our method consistently outperforms 20 existing approaches across both homophily and heterophily graphs, validating the effectiveness of leveraging natural relative similarity over artificial absolute similarity.

Figures

Figures reproduced from arXiv: 2505.05533 by the authors.

Figure 1
Figure 1. Visual vs. Graph contrastive learning: (a) image views [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Label consistency (the average proportion of neighbors [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. A philosophical comparison of absolute similarity and relative similarity in GCL. The right side presents the core idea of R [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Results of RELGCL with neighborhood range changing as k = 1, 2, 3, 4, respectively. sults. This highlights the advantage of modeling broader structural-semantic relationships over focusing solely on im￾mediate neighborhoods or artificial task designs. Performance on He…
Figure 5
Figure 5. Figure 5: Similarities between the anchor nodes’ and associated n-th [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Results of RELGCL with different threshold of α. in adapting to different similarity modeling strategies while maintaining effective relative similarity control. 5 Related Work Graph Contrastive Learning. GCL methods have followed the computer vision paradigm of maximi…
Figure 7
Figure 7. Figure 7: Results of RELGCL with neighborhood range changing as k = 1, 2, 3, 4, respectively. the average test accuracy with the corresponding standard de￾viation through 20 random initializations on each dataset. A few baselines do not use the public split of Cora, CiteSeer and…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Disentangled Multi-span Evolutionary Network against Temporal Knowledge Graph Reasoning

    cs.AI 2025-05 conditional novelty 5.0 of 10

    DiMNet combines multi-span cross-time message passing with disentangled active/stable node factors to set new state-of-the-art MRR on four TKG extrapolation benchmarks.

  2. Soft Graph Clustering for single-cell RNA Sequencing Data

    cs.LG 2025-07 conditional novelty 4.0 of 10

    A soft graph clustering framework combining ZINB autoencoding, joint normalized-cut embeddings, and optimal transport outperforms prior methods on ten scRNA-seq benchmarks.

  3. Distilling Closed-Source LLM's Knowledge for Locally Stable and Economic Biomedical Entity Linking

    cs.CL 2025-05 conditional novelty 4.0 of 10

    A closed-source LLM's re-ranking labels are distilled into a locally deployable open-source LLM, producing small but consistent Acc@1 gains in low-resource biomedical entity linking on two datasets.

Reference graph

Works this paper leans on

57 extracted references · 33 canonical work pages · cited by 3 Pith papers

  1. [1]

    Graph barlow twins: A self-supervised representation learning framework for graphs.Knowledge- Based Systems, 256:109631,

    [Bielak et al., 2022] Piotr Bielak, Tomasz Kajdanowicz, and Nitesh V Chawla. Graph barlow twins: A self-supervised representation learning framework for graphs.Knowledge- Based Systems, 256:109631,

    work page 2022

  2. [8]

    Momentum contrast for unsu- pervised visual representation learning

    [He et al., 2020] Kaiming He, Haoqi Fan, Yuxin Wu, Sain- ing Xie, and Ross Girshick. Momentum contrast for unsu- pervised visual representation learning. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 9729–9738,

    work page 2020

  3. [9]

    Hoffmann, Nadine Behrmann, Juergen Gall, Thomas Brox, and Mehdi Noroozi

    [Hoffmann et al., 2022] David T. Hoffmann, Nadine Behrmann, Juergen Gall, Thomas Brox, and Mehdi Noroozi. Ranking info noise contrastive estimation: Boosting contrastive learning via ranked positives. Pro- ceedings of the AAAI Conference on Artificial Intelligence, pages 897–905,

    work page 2022

  4. [10]

    Graphmae: Self-supervised masked graph autoencoders

    [Hou et al., 2022] Zhenyu Hou, Xiao Liu, Yukuo Cen, Yux- iao Dong, Hongxia Yang, Chun-Wei Wang, and Jie Tang. Graphmae: Self-supervised masked graph autoencoders. ArXiv, abs/2205.10803,

    arXiv 2022

  5. [11]

    Open graph benchmark: Datasets for machine learning on graphs

    [Hu et al., 2020] Weihua Hu, Matthias Fey, Marinka Zitnik, Yuxiao Dong, Hongyu Ren, Bowen Liu, Michele Catasta, and Jure Leskovec. Open graph benchmark: Datasets for machine learning on graphs. Advances in neural informa- tion processing systems, 33:22118–22133,

    2020

  6. [12]

    Automated self-supervised learning for graphs

    [Jin et al., 2021] Wei Jin, Xiaorui Liu, Xiangyu Zhao, Yao Ma, Neil Shah, and Jiliang Tang. Automated self-supervised learning for graphs. arXiv preprint arXiv:2106.05470,

    arXiv 2021

  7. [13]

    Supervised con- trastive learning

    [Khosla et al., 2020] Prannay Khosla, Piotr Teterwak, Chen Wang, Aaron Sarna, Yonglong Tian, Phillip Isola, Aaron Maschinot, Ce Liu, and Dilip Krishnan. Supervised con- trastive learning. Advances in neural information process- ing systems, 33:18661–18673,

    2020

  8. [14]

    Adam: A method for stochastic optimization

    [Kingma and Ba, 2014] Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980,

    arXiv 2014

Show all 57 references
  1. [17]

    Markov chains and mixing times , volume

    [Levin and Peres, 2017] David A Levin and Yuval Peres. Markov chains and mixing times , volume

    2017

  2. [19]

    Homogcl: Rethinking ho- mophily in graph contrastive learning

    [Li et al., 2023] Wen-Zhi Li, Chang-Dong Wang, Hui Xiong, and Jian-Huang Lai. Homogcl: Rethinking ho- mophily in graph contrastive learning. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Dis- covery and Data Mining, pages 1341–1352,

    2023

  3. [20]

    Prototypical graph contrastive learning

    [Lin et al., 2022] Shuai Lin, Chen Liu, Pan Zhou, Zi-Yuan Hu, Shuojia Wang, Ruihui Zhao, Yefeng Zheng, Liang Lin, Eric Xing, and Xiaodan Liang. Prototypical graph contrastive learning. IEEE transactions on neural net- works and learning systems, 35(2):2747–2758,

    2022

  4. [21]

    Random walks on graphs

    [Lovász, 1993] László Lovász. Random walks on graphs. Combinatorics, Paul erdos is eighty, 2(1-46):4,

    1993

  5. [23]

    Image-based recommendations on styles and substitutes

    [McAuley et al., 2015] Julian McAuley, Christopher Targett, Qinfeng Shi, and Anton Van Den Hengel. Image-based recommendations on styles and substitutes. In Proceed- ings of the 38th international ACM SIGIR conference on research and development in information retrieval , pages 43–52,

    2015

  6. [26]

    Simple unsupervised graph rep- resentation learning

    [Mo et al., 2022] Yujie Mo, Liang Peng, Jie Xu, Xiaoshuang Shi, and Xiaofeng Zhu. Simple unsupervised graph rep- resentation learning. Proceedings of the AAAI Conference on Artificial Intelligence, pages 7797–7805,

    2022

  7. [27]

    Lightcake: A lightweight framework for context-aware knowledge graph embed- ding

    [Ning et al., 2021] Zhiyuan Ning, Ziyue Qiao, Hao Dong, Yi Du, and Yuanchun Zhou. Lightcake: A lightweight framework for context-aware knowledge graph embed- ding. In Pacific-Asia Conference on Knowledge Discovery and Data Mining, pages 181–193. Springer,

    2021

  8. [28]

    Graph soft-contrastive learning via neighborhood ranking

    [Ning et al., 2022] Zhiyuan Ning, Pengfei Wang, Pengyang Wang, Ziyue Qiao, Wei Fan, Denghui Zhang, Yi Du, and Yuanchun Zhou. Graph soft-contrastive learning via neighborhood ranking. arXiv preprint arXiv:2209.13964,

    2022 arXiv

  9. [29]

    Fedgcs: A generative framework for ef- ficient client selection in federated learning via gradient- based optimization

    [Ning et al., 2024] Zhiyuan Ning, Chunlin Tian, Meng Xiao, Wei Fan, Pengyang Wang, Li Li, Pengfei Wang, and Yuanchun Zhou. Fedgcs: A generative framework for ef- ficient client selection in federated learning via gradient- based optimization. arXiv preprint arXiv:2405.06312 ,

    2024

  10. [30]

    Deep cut- informed graph embedding and clustering

    [Ning et al., 2025] Zhiyuan Ning, Zaitian Wang, Ran Zhang, Ping Xu, Kunpeng Liu, Pengyang Wang, Wei Ju, Pengfei Wang, Yuanchun Zhou, Erik Cambria, et al. Deep cut- informed graph embedding and clustering. arXiv preprint arXiv:2503.06635,

    2025 arXiv

  11. [33]

    Graph representation learning via graphi- cal mutual information maximization

    [Peng et al., 2020] Zhen Peng, Wenbing Huang, Minnan Luo, Qinghua Zheng, Yu Rong, Tingyang Xu, and Jun- zhou Huang. Graph representation learning via graphi- cal mutual information maximization. Proceedings of The Web Conference 2020,

    2020

  12. [34]

    Dual contrastive learning net- work for graph clustering

    [Peng et al., 2023] Xin Peng, Jieren Cheng, Xiangyan Tang, Jingxin Liu, and Jiahua Wu. Dual contrastive learning net- work for graph clustering. IEEE Transactions on Neural Networks and Learning Systems,

    2023

  13. [35]

    Glove: Global vec- tors for word representation

    [Pennington et al., 2014] Jeffrey Pennington, Richard Socher, and Christopher D Manning. Glove: Global vec- tors for word representation. In Proceedings of the 2014 conference on empirical methods in natural language processing (EMNLP), pages 1532–1543,

    2014

  14. [36]

    Character- izing graph datasets for node classification: Homophily- heterophily dichotomy and beyond

    [Platonov et al., 2024] Oleg Platonov, Denis Kuznedelev, Artem Babenko, and Liudmila Prokhorenkova. Character- izing graph datasets for node classification: Homophily- heterophily dichotomy and beyond. Advances in Neural Information Processing Systems, 36,

    2024

  15. [37]

    Context-enhanced entity and relation em- bedding for knowledge graph completion

    [Qiao et al., 2020] Ziyue Qiao, Zhiyuan Ning, Yi Du, and Yuanchun Zhou. Context-enhanced entity and relation em- bedding for knowledge graph completion. arXiv preprint arXiv:2012.07011,

    2020 arXiv

  16. [38]

    Rpt: toward transferable model on het- erogeneous researcher data via pre-training

    [Qiao et al., 2022] Ziyue Qiao, Yanjie Fu, Pengyang Wang, Meng Xiao, Zhiyuan Ning, Denghui Zhang, Yi Du, and Yuanchun Zhou. Rpt: toward transferable model on het- erogeneous researcher data via pre-training. IEEE Trans- actions on Big Data, 9(1):186–199,

    2022

  17. [39]

    A dual-channel semi-supervised learning framework on graphs via knowledge transfer and meta-learning

    [Qiao et al., 2024] Ziyue Qiao, Pengyang Wang, Pengfei Wang, Zhiyuan Ning, Yanjie Fu, Yi Du, Yuanchun Zhou, Jianqiang Huang, Xian-Sheng Hua, and Hui Xiong. A dual-channel semi-supervised learning framework on graphs via knowledge transfer and meta-learning. ACM Transactions on...

    2024

  18. [40]

    Conformity: a path-aware homophily mea- sure for node-attributed networks

    [Rossetti et al., 2021] Giulio Rossetti, Salvatore Citraro, and Letizia Milli. Conformity: a path-aware homophily mea- sure for node-attributed networks. IEEE Intelligent Sys- tems, 36(1):25–34,

    2021

  19. [41]

    Multi-scale attributed node embed- ding

    [Rozemberczki et al., 2021] Benedek Rozemberczki, Carl Allen, and Rik Sarkar. Multi-scale attributed node embed- ding. Journal of Complex Networks, 9(2):cnab014,

    2021

  20. [42]

    Collective classification in network data

    [Sen et al., 2008] Prithviraj Sen, Galileo Namata, Mustafa Bilgic, Lise Getoor, Brian Galligher, and Tina Eliassi-Rad. Collective classification in network data. AI magazine , 29(3):93–93,

    2008

  21. [44]

    Large- scale representation learning on graphs via bootstrapping

    [Thakoor et al., 2021] Shantanu Thakoor, Corentin Tallec, Mohammad Gheshlaghi Azar, Mehdi Azabou, Eva L Dyer, Remi Munos, Petar Veliˇckovi´c, and Michal Valko. Large- scale representation learning on graphs via bootstrapping. arXiv preprint arXiv:2102.06514,

    2021 arXiv

  22. [45]

    On mutual information maximization for representa- tion learning

    [Tschannen et al., 2019] Michael Tschannen, Josip Djo- longa, Paul K Rubenstein, Sylvain Gelly, and Mario Lu- cic. On mutual information maximization for representa- tion learning. arXiv preprint arXiv:1907.13625,

    2019 arXiv

  23. [47]

    Hamilton, Pietro Lio’, Yoshua Bengio, and R

    [Velickovicet al., 2019] Petar Velickovic, William Fedus, William L. Hamilton, Pietro Lio’, Yoshua Bengio, and R. Devon Hjelm. Deep graph infomax. ArXiv, abs/1809.10341,

    2019 arXiv

  24. [48]

    Single- pass contrastive learning can work for both homophilic and heterophilic graph

    [Wang et al., 2023] Haonan Wang, Jieyu Zhang, Qi Zhu, Wei Huang, Kenji Kawaguchi, and Xiaokui Xiao. Single- pass contrastive learning can work for both homophilic and heterophilic graph. Transactions on Machine Learn- ing Research,

    2023

  25. [49]

    Semi-supervised classification with graph convolutional networks

    [Welling and Kipf, 2016] Max Welling and Thomas N Kipf. Semi-supervised classification with graph convolutional networks. In J. International Conference on Learning Rep- resentations (ICLR 2017),

    2016

  26. [50]

    Simple and asymmetric graph contrastive learning without augmentations

    [Xiao et al., 2024] Teng Xiao, Huaisheng Zhu, Zhengyu Chen, and Suhang Wang. Simple and asymmetric graph contrastive learning without augmentations. Advances in Neural Information Processing Systems, 36,

    2024

  27. [51]

    sc- cdcg: Efficient deep structural clustering for single-cell rna-seq via deep cut-informed graph embedding

    [Xu et al., 2024] Ping Xu, Zhiyuan Ning, Meng Xiao, Gui- hai Feng, Xin Li, Yuanchun Zhou, and Pengfei Wang. sc- cdcg: Efficient deep structural clustering for single-cell rna-seq via deep cut-informed graph embedding. In Inter- national Conference on Database Systems for Advan...

    2024

  28. [52]

    From canonical correlation analysis to self-supervised graph neural networks

    [Zhang et al., 2021] Hengrui Zhang, Qitian Wu, Junchi Yan, David Wipf, and Philip S Yu. From canonical correlation analysis to self-supervised graph neural networks. In Ad- vances in Neural Information Processing Systems , pages 76–89,

    2021

  29. [53]

    Costa: Covariance- preserving feature augmentation for graph contrastive learning

    [Zhang et al., 2022] Yifei Zhang, Hao Zhu, Zixing Song, Piotr Koniusz, and Iriwn King. Costa: Covariance- preserving feature augmentation for graph contrastive learning. ArXiv, abs/2206.04726,

    2022 arXiv

  30. [54]

    To- ward graph self-supervised learning with contrastive ad- justed zooming

    [Zheng et al., 2022] Yizhen Zheng, Ming Jin, Shirui Pan, Yuan-Fang Li, Hao Peng, Ming Li, and Zhao Li. To- ward graph self-supervised learning with contrastive ad- justed zooming. IEEE Transactions on Neural Networks and Learning Systems,

    2022

  31. [55]

    Pathmlp: Smooth path towards high-order ho- mophily

    [Zhou et al., 2024] Jiajun Zhou, Chenxuan Xie, Shengbo Gong, Jiaxu Qian, Shanqing Yu, Qi Xuan, and Xiaoniu Yang. Pathmlp: Smooth path towards high-order ho- mophily. Neural Networks, 180:106650,

    2024

  32. [56]

    Liu, Shu Wu, and Liang Wang

    [Zhu et al., 2020b] Yanqiao Zhu, Yichen Xu, Feng Yu, Q. Liu, Shu Wu, and Liang Wang. Deep graph contrastive representation learning. ArXiv, abs/2006.04131,

    2006 arXiv

  33. [57]

    Liu, Shu Wu, and Liang Wang

    [Zhu et al., 2021] Yanqiao Zhu, Yichen Xu, Feng Yu, Q. Liu, Shu Wu, and Liang Wang. Graph contrastive learning with adaptive augmentation. Proceedings of the Web Confer- ence 2021, 2021

    2021

  34. [1993]

    Assessing the nonlinear decay of community similarity: Permutation and site-block resampling significance tests

    [Martínez-Santalla et al., 2022] Sara Martínez-Santalla, Ramiro Martín-Devasa, Carola Gómez-Rodríguez, Rosa M Crujeiras, and Andrés Baselga. Assessing the nonlinear decay of community similarity: Permutation and site-block resampling significance tests. Journal of Biogeography...

    2022

  35. [2001]

    End-to-end learning of visual representations from uncurated instructional videos

    [Miech et al., 2020] Antoine Miech, Jean-Baptiste Alayrac, Lucas Smaira, Ivan Laptev, Josef Sivic, and Andrew Zis- serman. End-to-end learning of visual representations from uncurated instructional videos. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pat- t...

    2020

  36. [2008]

    An overview of microsoft academic service (mas) and ap- plications

    [Sinha et al., 2015] Arnab Sinha, Zhihong Shen, Yang Song, Hao Ma, Darrin Eide, Bo-June Hsu, and Kuansan Wang. An overview of microsoft academic service (mas) and ap- plications. In Proceedings of the 24th international con- ference on world wide web, pages 243–246,

    2015

  37. [2012]

    Explor- ing simple siamese representation learning

    [Chen and He, 2021] Xinlei Chen and Kaiming He. Explor- ing simple siamese representation learning. In Proceed- ings of the IEEE/CVF conference on computer vision and pattern recognition, pages 15750–15758,

    2021

  38. [2014]

    Variational graph auto-encoders

    [Kipf and Welling, 2016] Thomas Kipf and Max Welling. Variational graph auto-encoders. ArXiv, abs/1611.07308,

    2016 arXiv

  39. [2015]

    Birds of a feather: Homophily in social networks.Annual review of sociology, pages 415– 444,

    [McPherson et al., 2001] Miller McPherson, Lynn Smith- Lovin, and James M Cook. Birds of a feather: Homophily in social networks.Annual review of sociology, pages 415– 444,

    2001

  40. [2016]

    Augmentation-free self-supervised learning on graphs

    [Lee et al., 2022] Namkyeong Lee, Junseok Lee, and Chany- oung Park. Augmentation-free self-supervised learning on graphs. Proceedings of the AAAI Conference on Artificial Intelligence, 36:7372–7380, Jun

    2022

  41. [2017]

    Graph communal contrastive learning

    [Li et al., 2022] Bolian Li, Baoyu Jing, and Hanghang Tong. Graph communal contrastive learning. In Proceedings of the ACM web conference 2022, pages 1203–1213,

    2022

  42. [2018]

    Geom-gcn: Geomet- ric graph convolutional networks

    [Pei et al., 2020] Hongbin Pei, Bingzhen Wei, Kevin Chen- Chuan Chang, Yu Lei, and Bo Yang. Geom-gcn: Geomet- ric graph convolutional networks. ArXiv, abs/2002.05287,

    2020 arXiv

  43. [2019]

    Graph attention networks

    [Velickovicet al., 2018] Petar Velickovic, Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Lio’, and Yoshua Bengio. Graph attention networks. ArXiv, abs/1710.10903,

    2018 arXiv

  44. [2020]

    Adaptive path-memory network for temporal knowledge graph reasoning

    [Dong et al., 2023] Hao Dong, Zhiyuan Ning, Pengyang Wang, Ziyue Qiao, Pengfei Wang, Yuanchun Zhou, and Yanjie Fu. Adaptive path-memory network for temporal knowledge graph reasoning. arXiv preprint arXiv:2304.12604,

    2023 arXiv

  45. [2021]

    A simple framework for contrastive learning of visual representations

    [Chen et al., 2020] Ting Chen, Simon Kornblith, Moham- mad Norouzi, and Geoffrey Hinton. A simple framework for contrastive learning of visual representations. In In- ternational conference on machine learning, pages 1597–

    2020

  46. [2022]

    Ho- mophily and long-run integration in social networks.Jour- nal of Economic Theory, 147(5):1754–1786,

    [Bramoullé et al., 2012] Yann Bramoullé, Sergio Currarini, Matthew O Jackson, Paolo Pin, and Brian W Rogers. Ho- mophily and long-run integration in social networks.Jour- nal of Economic Theory, 147(5):1754–1786,

    2012

  47. [2023]

    Tem- poral inductive path neural network for temporal knowl- edge graph reasoning

    [Dong et al., 2024] Hao Dong, Pengyang Wang, Meng Xiao, Zhiyuan Ning, Pengfei Wang, and Yuanchun Zhou. Tem- poral inductive path neural network for temporal knowl- edge graph reasoning. Artificial Intelligence, 329:104085,

    2024

  48. [2024]

    Contrastive multi-view repre- sentation learning on graphs

    [Hassani and Khasahmadi, 2020] Kaveh Hassani and Amir Hosein Khasahmadi. Contrastive multi-view repre- sentation learning on graphs. In International Conference on Machine Learning, pages 4116–4126. PMLR,

    2020

  49. [2025]

    Representation learning with contrastive predic- tive coding

    [Oord et al., 2018] Aaron van den Oord, Yazhe Li, and Oriol Vinyals. Representation learning with contrastive predic- tive coding. arXiv preprint arXiv:1807.03748,

    2018 arXiv

Pith tools

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