Pith. sign in

REVIEW 4 major objections 5 minor 77 references

Unified Graph Networks (UGN): A Deep Neural Framework for Solving Graph Problems

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

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

Pith's one-line read A single encoder-decoder graph network, built from a GCN encoder and a Conv2D decoder over pairwise node-latent products, claims top results on ten of twelve datasets spanning six graph tasks.

desk verdict A useful engineering combination, but the paper's central empirical claim is contradicted by its own tables and the HCP margin is likely an unreported mean-target baseline artifact. read the letter →

arxiv 2502.07500 v1 pith:XR2SLHZH submitted 2025-02-11 cs.LG

Rudrajit Dawn , Madhusudan Ghosh , Partha Basuchowdhuri , Sudip Kumar Naskar This is my paper · ORCID
classification cs.LG
keywords UnifiedGraphNetworkgraph-to-graphtranslationlinkpredictioncommunitydetectionknowledgecompletionconvolutionalsupernodefeaturemeantargetconnectivitymatrix
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 proposes the Unified Graph Network (UGN), a single encoder-decoder architecture intended to serve as a general-purpose model for graph problems, analogous to how transformers serve text and images. Graph-convolution layers in the encoder turn each node into a latent vector, and a two-dimensional convolutional decoder classifies nodes, edges, or graph-level outputs from the pairwise product of latent vectors. The framework is tested on six graph tasks and twelve datasets, with only small task-specific additions: supernode features for graphs with no node attributes, an unsupervised loss for semi-supervised community detection, and a mean target connectivity matrix for complete-graph translation. The paper reports that UGN outperforms existing specialized models on ten of the twelve datasets and is comparable on the rest. If the results hold, one relatively small framework could replace many task-specific graph models.

What carries the argument

The load-bearing mechanism is the intermediate product matrix: for an ordered node pair $(u,v)$, the encoder's latent vectors are multiplied as $M = l_v^T l_u$, exposing every cross-feature interaction between the two nodes to the decoder at once; this $L \times L$ matrix is then processed like a one-channel image by stacked Conv2D, max-pooling, and fully connected layers. Around that core sit three named components. Supernode features coarsen a large graph into mutually exclusive node subsets and encode each node by normalized connection counts to those subsets. An unsupervised mean-squared-error loss $\mathcal{L}_u = \sum_{(u,v)\in E} \frac{1}{c}\sum_i (O[u][i]-O[v][i])^2$ pushes connected nodes toward the same class distribution in semi-supervised tasks. The mean target connectivity matrix $\bar{T}[i][j] = \frac{1}{|D|}\sum_{T\in D} T[i][j]$ is subtracted from each complete-graph target matrix so that the model learns to predict only the residual difference from the dataset average.

What would settle it

Retrain UGN on the HCP data without MTCM and also evaluate a trivial baseline that predicts the training-set average functional connectivity matrix for every test subject; if the average baseline matches or exceeds UGN's reported Pearson correlations of 0.61 to 0.68, then the learned structural-to-functional mapping is not the source of the reported gain.

Watch

Extended reading notes

Core claim

The central claim, stated on the paper's own terms, is that the same GCN-to-Conv2D encoder-decoder stack solves node classification, edge classification, link prediction, community detection, graph-to-graph translation, and knowledge graph completion with minimal task-specific changes. The decoder forms the product matrix $M = l_v^T l_u$ from the latent vectors of two nodes and processes all pairwise feature interactions as a one-channel image through convolution, max-pooling, and linear layers. With supernode features, an unsupervised loss term, and the mean target connectivity matrix as add-ons, the paper reports state-of-the-art or comparable accuracy on IoT malware-confinement graphs, chemistry reaction prediction, Epinions and Slashdot link prediction, three community-detection benchmarks, the Human Connectome Project brain-translation task, three biomedical interaction datasets, and YAGO knowledge graph completion.

Load-bearing premise

The load-bearing premise is that the training-set average target connectivity matrix is a legitimate prior to subtract from every test subject in the brain-translation task, and that the residual individual variation is what the model actually learns from structural connectivity; the paper never reports how a model that simply outputs the average target matrix for everyone would compare.

Editorial extensions

If this is right

  • If UGN's reported numbers reproduce, one architecture with small add-ons can replace separate models for node classification, edge classification, link prediction, community detection, graph-to-graph translation, and knowledge graph completion.
  • The supernode feature makes large featureless graphs (Epinions and Slashdot, with 22k–82k nodes) trainable without one-hot node features, raising Slashdot edge accuracy from 0.80 to 0.82.
  • The unsupervised squared-difference loss lets the same supervised architecture handle semi-supervised community detection, reaching 1.00 node accuracy on Zachary's karate club and 0.92 on American College football.
  • The MTCM treatment reduces complete-graph translation to residual learning, and without it UGN's HCP score falls to 0.36 Pearson correlation, the level of the previous baseline.
  • On YAGO knowledge graph completion, the paper reports that about 156,000 training edges are enough to beat the previous best HITS@1 (0.53 versus 0.51), and full training reaches 0.61.

Reading between the lines

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

  • The HCP ablation invites an inference the paper does not draw: since removing MTCM drops UGN to the previous baseline, a large share of the reported brain-translation gain may come from predicting the training-set average functional connectivity matrix rather than from the learned structural-to-functional mapping; a trivial average-target baseline would settle this.
  • Because the decoder reads the outer product of two latent vectors, it is a bilinear pairwise readout rather than a graph-specific mechanism, so the same architecture could be applied to other pairwise structured-prediction problems such as drug-target binding, question-answer ranking, or recommender scoring.
  • The zero-shot results do not say whether the supernode partition itself matters; comparing random, degree-based, and learned partitions would separate the effect of coarse connectivity statistics from the effect of the particular coarsening scheme.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes UGN, an encoder-decoder framework that combines graph convolutional layers with two-dimensional convolutional layers, and applies it to six graph-learning tasks: graph translation (IoT, Chemistry), community detection (Karate, Football, Books), link prediction (Epinions, Slashdot, DDI, DTI, PPI), connectivity prediction (HCP), and knowledge graph completion (YAGO). Three task-specific extensions are introduced: supernode-based node features for large graphs, an unsupervised loss for semi-supervised settings, and a mean target connectivity matrix (MTCM) for complete graph translation. The paper claims that UGN outperforms state-of-the-art baselines on ten datasets and is comparable on the remaining dataset(s). The manuscript includes ablation studies for the intermediate-matrix decoder, supernode features, MTCM, and the unsupervised loss.

Significance. If the claims were fully supported, a single framework that achieves state-of-the-art or near-state-of-the-art results across six heterogeneous graph tasks would be a useful contribution, especially because the authors provide an anonymous code link and evaluate on a wide range of datasets. The conceptual idea of unifying tasks through a GCN-encoder and image-style decoder is plausible, and the ablations show that the proposed components affect performance. However, the empirical contribution as stated is not established: the paper's own tables contradict the headline 'ten datasets' claim, the largest reported margin (HCP) lacks a critical baseline, and no error bars or significance tests are provided anywhere. The claimed significance is therefore currently unsupported.

major comments (4)
  1. [Abstract, Section 8 Conclusion, Tables 1b, 1c, 1f]
  2. [Section 3.2.3, Eq. (7)-(8); Section 5.5.1; Figure 2c]
  3. [All result tables (Tables 1a-1f, 3, 4, 5)]
  4. [Section 3.2.1 and Section 5.4.2]
minor comments (5)
  1. [Section 3.2.3 heading]
  2. [Section 3.2.2]
  3. [Section 5.1]
  4. [Table 2]
  5. [Eq. (7)]

Circularity Check

1 steps flagged · score 4.0 of 10

HCP 'significant margin' is dominated by the fitted MTCM training-set mean: UGN without MTCM scores 0.36, exactly matching GT-GAN, so the flagship brain-connectivity result largely reduces to the fitted constant.

  1. fitted input called prediction [Section 3.2.3 (Eqs. 7-8), Section 5.5.1, Figure 2c ablation]
    "The MTCM is constructed using equation (7) where T represents the required MTCM... The predicted connectivity matrix is obtained by adding the MTCM to the predicted difference matrix from a source connectivity matrix. Using this method, the model can exploit the similarity among the target connectivity matrices in its prediction and helps improve the evaluation scores. /// ...we achieved an average pearson correlation of 0.36 for the motoring task on the HCP dataset... Thus, our baseline architecture without the MTCM component performs at par with the previous SOTA model, GT-GAN"

    MTCM (Eq. 7) is the training-set mean of the target FC matrices — a fitted statistic of the very quantity being 'predicted.' The test-time output is, by construction, MTCM plus a learned residual (Eq. 8 with the add-back), and the reported HCP correlations (0.61-0.68 vs. prior SOTA 0.34-0.45) credit the whole sum. The paper's own ablation (Figure 2c) shows that removing MTCM drops the motor-task score to 0.36, exactly GT-GAN's 0.36, so the learned SC-to-FC mapper alone equals the previous SOTA; the advertised margin therefore reduces, by the paper's own equations, to the fitted training-mean constant, and the mean-only baseline isolating the network's true contribution is never reported. The prediction is thus largely forced by the fitted input.

full rationale

This is an empirical architecture paper, not a derivation, and most of it is self-contained: there is no self-citation chain (the reference list is entirely external), no imported uniqueness theorem, and no ansatz smuggled in by citation; the GCN encoder, pairwise-multiplication intermediate matrix, and Conv2D decoder are a genuine construction rather than a renaming of a known method. The one circularity-adjacent step is the MTCM component for the HCP graph-translation task (Section 3.2.3). There, the training-set mean of the target FC matrices is subtracted from every target (Eq. 8) and added back at inference, so the forecast literally equals a fitted constant plus a learned residual, and the paper's own ablation demonstrates that the constant is what produces the SOTA margin: without MTCM, UGN scores 0.36 on the motor task, identical to GT-GAN, while with MTCM it scores 0.66. The HCP 'significant margin' therefore reduces, by the paper's own equations and Figure 2c, to the fitted target mean, and no mean-only baseline is reported to measure the residual mapper's independent contribution. This is a partial fitted-input-called-prediction circularity affecting the flagship claim. The remaining results (IoT, community detection, Epinions/Slashdot, DTI/PPI, YAGO) are standard supervised predictions against external baselines, with no fitting of target statistics; the YAGO left/right one-hot features encode training-set relation membership, which is legitimate feature engineering, and the supernode features, unsupervised loss, and intermediate-matrix ablations are all empirically evaluated rather than assumed. Separately, for correctness (not circularity), the abstract's claim of a 'significant margin on ten datasets' is inconsistent with the conclusion's 'two datasets' and with Tables 1b/1f, where UGN trails NEC-DGT on Chemistry and HOGCN on DDI and ties on Karate and DTI, with no significance tests reported anywhere. On balance, one flagship result substantially reduces to a fitted statistic while the central framework retains independent content, giving a score of 4.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical or mathematical entities. The load-bearing assumptions are the community smoothness prior, the common-mean target prior in MTCM, and the efficacy of pairwise-multiplication features. The free parameters are hand-selected architecture sizes, supernode counts, random vector dimensions, and the training-set average target matrix. None of these are derived or fitted with a reported search procedure.

free parameters (6)
  • Latent dimension L = not reported per dataset
    The size of the node latent vectors from the GCN encoder; the paper says k, s, t range from three to five depending on dataset but does not list exact values.
  • Number of GCN layers k = 3-5
    Chosen by hand per dataset; affects encoder capacity and is not tuned systematically.
  • Number of conv+pool blocks s = 3-5
    Depth of the 2D convolutional decoder, selected by hand.
  • Supernode count for Slashdot = 83
    82 supernodes of 1000 nodes plus one of 168; used to build the 176-dimensional node feature vector.
  • Random vector dimension = 10 (Slashdot), 20 (YAGO)
    Random augmentation added to break ties in edge-count and one-hot features; introduces nondeterminism not controlled by a seed.
  • MTCM mean target matrix = average FC matrix over the training set per HCP task
    Computed from training target matrices and added to the predicted residual; a fitted statistic that drives most of the HCP performance gain.
assumptions (4)
  • standard math GCN propagation rule uses symmetrically normalized adjacency with self-loops (Eq. 1-2).
    Standard graph convolutional layer from Kipf and Welling, assumed without proof.
  • domain assumption Densely connected nodes tend to belong to the same community, so the unsupervised MSE loss over edges is a valid regularizer.
    Stated in Section 3.2.2; this is the graph smoothness prior underpinning L_u.
  • domain assumption Target connectivity matrices share a common mean, so subtracting the training-set average leaves a predictable residual.
    Stated in Section 3.2.3; without this, MTCM would not improve HCP prediction.
  • domain assumption Pairwise product of latent vectors captures inter-feature dependencies that concatenation does not.
    Asserted in Section 3.1 around Eq. (4); not proven or empirically isolated.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Unified Graph Networks (UGN): A Deep Neural Framework for Solving Graph Problems." pith.science (2026). https://pith.science/paper/XR2SLHZH

@misc{pith2026250207500,
  author       = {Pith},
  title        = {Pith review of: Unified Graph Networks (UGN): A Deep Neural Framework for Solving Graph Problems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XR2SLHZH}},
  note         = {Machine review of arXiv:2502.07500}
}
read the original abstract

Deep neural networks have enabled researchers to create powerful generalized frameworks, such as transformers, that can be used to solve well-studied problems in various application domains, such as text and image. However, such generalized frameworks are not available for solving graph problems. Graph structures are ubiquitous in many applications around us and many graph problems have been widely studied over years. In recent times, there has been a surge in deep neural network based approaches to solve graph problems, with growing availability of graph structured datasets across diverse domains. Nevertheless, existing methods are mostly tailored to solve a specific task and lack the capability to create a generalized model leading to solutions for different downstream tasks. In this work, we propose a novel, resource-efficient framework named \emph{U}nified \emph{G}raph \emph{N}etwork (UGN) by leveraging the feature extraction capability of graph convolutional neural networks (GCN) and 2-dimensional convolutional neural networks (Conv2D). UGN unifies various graph learning tasks, such as link prediction, node classification, community detection, graph-to-graph translation, knowledge graph completion, and more, within a cohesive framework, while exercising minimal task-specific extensions (e.g., formation of supernodes for coarsening massive networks to increase scalability, use of \textit{mean target connectivity matrix} (MTCM) representation for achieving scalability in graph translation task, etc.) to enhance the generalization capability of graph learning and analysis. We test the novel UGN framework for six uncorrelated graph problems, using twelve different datasets. Experimental results show that UGN outperforms the state-of-the-art baselines by a significant margin on ten datasets, while producing comparable results on the remaining dataset.

Figures

Figures reproduced from arXiv: 2502.07500 by the authors.

Figure 1
Figure 1. Pictorial overview of our UGN framework. 3 MODEL ARCHITECTURE 3.1 Encoder-Decoder Model The proposed model uses an encoder-decoder architecture. The encoder transforms the input into a dense vector that captures key features, typically with lower dimensions to reduce storage and computational costs. The decoder then uses this dense representa￾tion to generate the desired output in the appropriate format.The overview… view at source ↗
Figure 2
Figure 2. Ablation study by component replacement and removal on (a) IoT-60 dataset, (b) Slashdot dataset, (c) HCP dataset Motoring task, and (d) Zachary’s karate club dataset dimension is to be decreased, then the random vector dimension is reduced. The results of the zero-shot experiments are reported in [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

77 extracted references · 69 canonical work pages

  1. [1]

    Abdelnour, H

    F. Abdelnour, H. U. Voss, and A. Raj. Network diffusion accurately models the relationship between structural and functional brain connectivity networks. Neuroimage, 90:335–347, 2014

    work page 2014

  2. [2]

    Abdelnour, M

    F. Abdelnour, M. Dayan, O. Devinsky, T. Thesen, and A. Raj. Functional brain connectivity is predictable from anatomic network’s laplacian eigen-structure. NeuroImage, 172:728–739, 2018

    2018

  3. [3]

    Abu-El-Haija, B

    S. Abu-El-Haija, B. Perozzi, and R. Al-Rfou. Learning edge representations via low- rank asymmetric projections. In Proceedings of the 2017 ACM on Conference on Information and Knowledge Management, pages 1787–1796, 2017

    work page 2017

  4. [4]

    L. Bai, L. Yao, C. Li, X. Wang, and C. Wang. Adaptive graph convolutional recur- rent network for traffic forecasting. Advances in neural information processing systems, 33:17804–17815, 2020

    work page 2020

  5. [5]

    Balažević, C

    I. Balažević, C. Allen, and T. Hospedales. Tucker: Tensor factorization for knowledge graph completion. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 5185– 5194, 2019

    work page 2019

  6. [6]

    Battaglia, R

    P. Battaglia, R. Pascanu, M. Lai, D. J. Rezende, and K. kavukcuoglu. Interaction networks for learning about objects, relations and physics. In Proceedings of the 30th International Conference on Neural Information Processing Systems, pages 4509–4517, 2016

    work page 2016

  7. [7]

    Bordes, N

    A. Bordes, N. Usunier, A. Garcia-Durán, J. Weston, and O. Yakhnenko. Translat- ing embeddings for modeling multi-relational data. In Proceedings of the 26th International Conference on Neural Information Processing Systems-Volume 2, pages 2787–2795, 2013

    work page 2013

  8. [8]

    Bruna, W

    J. Bruna, W. Zaremba, A. Szlam, and Y. Lecun. Spectral networks and lo- cally connected networks on graphs. In International Conference on Learning Representations (ICLR2014), CBLS, April 2014, 2014

    work page 2014

Show all 77 references
  1. [9]

    B. Cai, L. Zeng, Y. Wang, H. Li, and Y. Hu. Community detection method based on node density, degree centrality, and k-means clustering in complex network. Entropy, 21(12):1145, 2019

    2019

  2. [10]

    S. Cao, W. Lu, and Q. Xu. Grarep: Learning graph representations with global structural information. In Proceedings of the 24th ACM international on conference on information and knowledge management, pages 891–900, 2015

    2015

  3. [11]

    Chami, A

    I. Chami, A. Wolf, D.-C. Juan, F. Sala, S. Ravi, and C. Ré. Low-dimensional hyper- bolic knowledge graph embeddings. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 6901–6914, 2020

    2020

  4. [12]

    A. Clauset. Finding local community structure in networks. Physical review E, 72(2):026132, 2005

    2005

  5. [13]

    Clauset, M

    A. Clauset, M. E. Newman, and C. Moore. Finding community structure in very large networks. Physical review E, 70(6):066111, 2004

    2004

  6. [14]

    Z. Z. Darban and M. H. Valipour. Ghrs: Graph-based hybrid recommenda- tion system with application to movie recommendation. Expert Systems with Applications, 200:116850, 2022

    2022

  7. [15]

    Defferrard, X

    M. Defferrard, X. Bresson, and P. Vandergheynst. Convolutional neural net- works on graphs with fast localized spectral filtering. In Proceedings of the 30th International Conference on Neural Information Processing Systems, pages 3844–3852, 2016

    2016

  8. [16]

    Dettmers, P

    T. Dettmers, P. Minervini, P. Stenetorp, and S. Riedel. Convolutional 2d knowledge graph embeddings. In 32nd AAAI Conference on Artificial Intelligence, AAAI 2018, volume 32, pages 1811–1818. AAI Publications, 2018

    2018

  9. [17]

    Fukushima

    K. Fukushima. Neocognitron: A self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in position. Biological cybernetics, 36(4):193–202, 1980

    1980

  10. [18]

    R. F. Galán. On how network architecture determines the dominant patterns of spontaneous neural activity. PloS one, 3(5):e2148, 2008

    2008

  11. [19]

    Girvan and M

    M. Girvan and M. E. Newman. Community structure in social and biological networks. Proceedings of the national academy of sciences, 99(12):7821–7826, 2002

    2002

  12. [20]

    Grover and J

    A. Grover and J. Leskovec. node2vec: Scalable feature learning for networks. In Proceedings of the 22nd ACM SIGKDD international conference on Knowledge discovery and data mining, pages 855–864, 2016

    2016

  13. [21]

    X. Guo, L. Zhao, C. Nowzari, S. Rafatirad, H. Homayoun, and S. M. P. Dinakarrao. Deep multi-attributed graph translation with node-edge co-evolution. In 2019 IEEE International Conference on Data Mining (ICDM), pages 250–259. IEEE, 2019

    2019

  14. [22]

    X. Guo, L. Wu, and L. Zhao. Deep graph translation. IEEE Transactions on Neural Networks and Learning Systems, pages 1–10, 2022. doi: 10.1109/TNNLS. 2022.3144670

    2022

  15. [23]

    W. L. Hamilton, R. Ying, and J. Leskovec. Representation learning on graphs: Methods and applications. IEEE Data Eng. Bull., 40(3):52–74, 2017. URL http: //sites.computer.org/debull/A17sept/\p52.pdf

    2017

  16. [24]

    Huang, C

    K. Huang, C. Xiao, L. M. Glass, M. Zitnik, and J. Sun. Skipgnn: predicting molecular interactions with skip-graph networks. Scientific reports, 10(1):1–16, 2020

    2020

  17. [25]

    W. Jin, C. W. Coley, R. Barzilay, and T. Jaakkola. Predicting organic reaction out- comes with weisfeiler-lehman network. In Proceedings of the 31st International Conference on Neural Information Processing Systems, pages 2604–2613, 2017

    2017

  18. [26]

    S. M. Kazemi and D. Poole. Simple embedding for link prediction in knowl- edge graphs. In Proceedings of the 32nd International Conference on Neural Information Processing Systems, pages 4289–4300, 2018

    2018

  19. [27]

    T. N. Kipf and M. Welling. Variational graph auto-encoders. arXiv preprint arXiv:1611.07308, 2016

    2016 arXiv

  20. [28]

    T. N. Kipf and M. Welling. Semi-supervised classification with graph convolu- tional networks. In 5th International Conference on Learning Representations, ICLR 2017, Toulon, France, April 24-26, 2017, Conference Track Proceedings. OpenReview.net, 2017. URL https://openreview....

    2017

  21. [29]

    Kishan, R

    K. Kishan, R. Li, F. Cui, and A. R. Haake. Predicting biomedical interactions with higher-order graph convolutional networks. IEEE/ACM transactions on computational biology and bioinformatics, 19(2):676–687, 2021

    2021

  22. [30]

    T. G. Kolda and B. W. Bader. Tensor decompositions and applications. SIAM review, 51(3):455–500, 2009

    2009

  23. [31]

    I. A. Kovács, K. Luck, K. Spirohn, Y. Wang, C. Pollis, S. Schlabach, W. Bian, D.-K. Kim, N. Kishore, T. Hao, et al. Network-based prediction of protein interactions. Nature Communications, 10(1):1–8, 2019

    2019

  24. [32]

    V. Krebs. Books about us politics. unpublished, http://www. orgnet. com, 2004

    2004

  25. [33]

    Lancichinetti and S

    A. Lancichinetti and S. Fortunato. Limits of modularity maximization in commu- nity detection. Physical review E, 84(6):066122, 2011

    2011

  26. [34]

    Leskovec, K

    J. Leskovec, K. J. Lang, A. Dasgupta, and M. W. Mahoney. Community structure in large networks: Natural cluster sizes and the absence of large well-defined clusters. Internet Mathematics, 6(1):29–123, 2009

    2009

  27. [35]

    J. Li, J. Zhu, and B. Zhang. Discriminative deep random walk for network classification. In Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1004–1013, 2016

    2016

  28. [36]

    Y. Li, R. Yu, C. Shahabi, and Y. Liu. Diffusion convolutional recurrent neural net- work: Data-driven traffic forecasting. In International Conference on Learning Representations, 2018. URL https://openreview.net/forum?id=SJiHXGWAZ

    2018

  29. [37]

    D. Lowe. Patent reaction extraction: downloads, 2014

    2014

  30. [38]

    Z. Lu, J. Wahlström, and A. Nehorai. Community detection in complex networks via clique conductance. Scientific reports, 8(1):5982, 2018

    2018

  31. [39]

    Luck, D.-K

    K. Luck, D.-K. Kim, L. Lambourne, K. Spirohn, B. E. Begg, W. Bian, R. Brignall, T. Cafarelli, F. J. Campos-Laborie, B. Charloteaux, et al. A reference map of the human binary protein interactome. Nature, 580(7803):402–408, 2020

    2020

  32. [40]

    Mahmood and M

    A. Mahmood and M. Small. Subspace based network community detection using sparse linear coding. IEEE Transactions on Knowledge and Data Engineering, 28(3):801–812, 2015

    2015

  33. [41]

    Marinka Zitnik, R

    S. Marinka Zitnik, R. Sosič, and J. Leskovec. Biosnap datasets: Stanford biomedical network dataset collection. URL http://snap. stanford. edu/biodata, 2018

    2018

  34. [42]

    Meier, P

    J. Meier, P. Tewarie, A. Hillebrand, L. Douw, B. W. Van Dijk, S. M. Stufflebeam, and P. Van Mieghem. A mapping between structural and functional brain networks. Brain connectivity, 6(4):298–311, 2016

    2016

  35. [43]

    Nebli, U

    A. Nebli, U. A. Kaplan, and I. Rekik. Deep evographnet architecture for time- dependent brain graph data synthesis from a single timepoint. In International Workshopon PRedictive Intelligence In MEdicine, pages 144–155. Springer, 2020

    2020

  36. [44]

    M. E. Newman. Finding community structure in networks using the eigenvectors of matrices. Physical review E, 74(3):036104, 2006

    2006

  37. [45]

    M. E. Newman. Spectral methods for community detection and graph partitioning. Physical Review E, 88(4):042822, 2013

    2013

  38. [46]

    Nghiep Tran and A

    H. Nghiep Tran and A. Takasu. Multi-partition embedding interaction with block term format for knowledge graph completion. In ECAI 2020, pages 833–840. IOS Press, 2020

    2020

  39. [47]

    Perozzi, R

    B. Perozzi, R. Al-Rfou, and S. Skiena. Deepwalk: Online learning of social repre- sentations. In Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining, pages 701–710, 2014. Unified Graph Networks (UGN): A Deep Neural Framework for S...

    2014

  40. [48]

    Reichardt and S

    J. Reichardt and S. Bornholdt. Statistical mechanics of community detection. Physical review E, 74(1):016110, 2006

    2006

  41. [49]

    Y. Rong, Y. Bian, T. Xu, W. Xie, Y. Wei, W. Huang, and J. Huang. Self- supervised graph transformer on large-scale molecular data. Advances in Neural Information Processing Systems, 33:12559–12571, 2020

    2020

  42. [50]

    Rosvall and C

    M. Rosvall and C. T. Bergstrom. Maps of random walks on complex networks reveal community structure. Proceedings of the national academy of sciences, 105(4):1118–1123, 2008

    2008

  43. [51]

    Scarselli, M

    F. Scarselli, M. Gori, A. C. Tsoi, M. Hagenbuchner, and G. Monfardini. The graph neural network model. IEEE Transactions on Neural Networks, 20(1):61–80, 2009. doi: 10.1109/TNN.2008.2005605

    2009

  44. [52]

    Shang, J

    R. Shang, J. Bai, L. Jiao, and C. Jin. Community detection based on modularity and an improved genetic algorithm. Physica A: Statistical Mechanics and its Applications, 392(5):1215–1231, 2013

    2013

  45. [53]

    Simonovsky and N

    M. Simonovsky and N. Komodakis. Graphvae: Towards generation of small graphs using variational autoencoders. In Artificial Neural Networks and Machine Learning–ICANN 2018: 27th International Conference on Artificial Neural Networks, Rhodes, Greece, October 4-7, 2018, Proceedin...

    2018

  46. [54]

    F. M. Suchanek, G. Kasneci, and G. Weikum. Yago: a core of semantic knowledge. In Proceedings of the 16th international conference on World Wide Web, pages 697–706, 2007

    2007

  47. [55]

    Sun, Z.-H

    Z. Sun, Z.-H. Deng, J.-Y. Nie, and J. Tang. Rotate: Knowledge graph embedding by relational rotation in complex space. In International Conference on Learning Representations, 2019

    2019

  48. [56]

    J. Tang, H. Gao, H. Liu, and A. D. Sarma. eTrust: Understanding trust evolution in an online world. In KDD, 2012

    2012

  49. [57]

    J. Tang, M. Qu, M. Wang, M. Zhang, J. Yan, and Q. Mei. Line: Large-scale informa- tion network embedding. In Proceedings of the 24th international conference on world wide web, pages 1067–1077, 2015

    2015

  50. [58]

    K. Tang, Y. Niu, J. Huang, J. Shi, and H. Zhang. Unbiased scene graph generation from biased training. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 3716–3725, 2020

    2020

  51. [59]

    Y. Tang, J. Huang, G. Wang, X. He, and B. Zhou. Orthogonal relation transforms with graph context modeling for knowledge graph embedding. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 2713–2722, 2020

    2020

  52. [60]

    Tian and W

    B. Tian and W. Li. Community detection method based on mixed-norm sparse subspace clustering. Neurocomputing, 275:2150–2161, 2018

    2018

  53. [61]

    H. N. Tran and A. Takasu. Analyzing knowledge graph embedding methods from a multi-embedding interaction perspective. Representations, 11:12, 2019

    2019

  54. [62]

    H. N. Tran and A. Takasu. MEIM: multi-partition embedding interaction beyond block term format for efficient and expressive link prediction. In L. D. Raedt, edi- tor, Proceedings of the Thirty-First International Joint Conference on Artificial Intelligence, IJCAI 2022, Vienna,...

    2022 doi

  55. [63]

    Trouillon, J

    T. Trouillon, J. Welbl, S. Riedel, É. Gaussier, and G. Bouchard. Complex em- beddings for simple link prediction. In International conference on machine learning, pages 2071–2080. PMLR, 2016

    2016

  56. [64]

    C. Tu, W. Zhang, Z. Liu, M. Sun, et al. Max-margin deepwalk: Discriminative learning of network representation. In IJCAI, volume 2016, pages 3889–3895, 2016

    2016

  57. [65]

    D. C. Van Essen, S. M. Smith, D. M. Barch, T. E. Behrens, E. Yacoub, K. Ugurbil, W.- M. H. Consortium, et al. The wu-minn human connectome project: an overview. Neuroimage, 80:62–79, 2013

    2013

  58. [66]

    Vashishth, S

    S. Vashishth, S. Sanyal, V. Nitin, N. Agrawal, and P. Talukdar. Interacte: Im- proving convolution-based knowledge graph embeddings by increasing feature interactions. In Proceedings of the AAAI conference on artificial intelligence, volume 34, number 03, pages 3009–3016, 2020

    2020

  59. [67]

    Vashishth, S

    S. Vashishth, S. Sanyal, V. Nitin, and P. Talukdar. Composition-based multi- relational graph convolutional networks. In International Conference on Learning Representations, 2020

    2020

  60. [68]

    C. Wang, C. Wang, Z. Wang, X. Ye, and P. S. Yu. Edge2vec: Edge-based social network embedding. ACM Transactions on Knowledge Discovery from Data (TKDD), 14(4):1–24, 2020

    2020

  61. [69]

    D. Wang, P. Cui, and W. Zhu. Structural deep network embedding. InProceedings of the 22nd ACM SIGKDD international conference on Knowledge discovery and data mining, pages 1225–1234, 2016

    2016

  62. [70]

    Z. Wang, X. Ye, C. Wang, Y. Wu, C. Wang, and K. Liang. Rsdne: explor- ing relaxed similarity and dissimilarity from completely-imbalanced labels for network embedding. In Proceedings of the Thirty-Second AAAI Conference on Artificial Intelligence and Thirtieth Innovative Appli...

    2018

  63. [71]

    Y. Wu, M. Cashman, N. Choma, É. T. Prates, V. Vergara, M. Shah, A. Chen, A. Clyde, T. Brettin, W. de Jong, et al. Spatial graph attention and curiosity-driven policy for antiviral drug discovery. In10th International Conference on Learning Representations, ICLR, 2022

    2022

  64. [72]

    Xu and R

    C. Xu and R. Li. Relation embedding with dihedral group in knowledge graph. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 263–272, 2019

    2019

  65. [73]

    J. You, R. Ying, X. Ren, W. Hamilton, and J. Leskovec. Graphrnn: Generating realistic graphs with deep auto-regressive models. In International conference on machine learning, pages 5708–5717. PMLR, 2018

    2018

  66. [74]

    B. Yu, H. Yin, and Z. Zhu. Spatio-temporal graph convolutional networks: A deep learning framework for traffic forecasting. In International Joint Conference on Artificial Intelligence, pages 3634–3640, 07 2018. doi: 10.24963/ijcai.2018/505

    2018 doi

  67. [75]

    W. W. Zachary. An information flow model for conflict and fission in small groups. Journal of anthropological research, 33(4):452–473, 1977

    1977

  68. [76]

    Zhang, Y

    S. Zhang, Y. Tay, L. Yao, and Q. Liu. Quaternion knowledge graph embeddings. In Proceedings of the 33rd International Conference on Neural Information Processing Systems, pages 2735–2745, 2019

    2019

  69. [77]

    Zhang, J

    Z. Zhang, J. Cai, and J. Wang. Duality-induced regularizer for tensor factorization based knowledge graph completion. Advances in Neural Information Processing Systems, 33:21604–21615, 2020

    2020

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.