Pith. sign in

REVIEW 4 major objections 5 minor 38 references

Graph Embedding Based Hybrid Social Recommendation System

T0 review · 4 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Pith's one-line read A hybrid of three social-graph embeddings more than doubles recommendation coverage over the best single embedding on the Yelp graph.

desk verdict The hybrid's edge over single embeddings is an artifact of training and evaluating on the same 100 users; the paper is a clear but empirically unsupported incremental study. read the letter →

arxiv 1908.09454 v1 pith:RNEF7UXT submitted 2019-08-26 cs.SI cs.IR

classification cs.SIcs.IR
keywords GraphEmbeddingSocialnetworksRecommendationSystemSpectralClusteringhybridmodelsnode2vecHOPE
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper sets out to show that social circles alone, encoded as graph embeddings, can drive item recommendations, and that a neural hybrid of three embeddings beats each embedding used alone. The authors build an implicit weighted user graph from Yelp friendships and similar restaurant ratings, embed it three ways that preserve different structural aspects, and train a small dense network to decide, per user and per restaurant, which embeddings' recommendations to trust. On 100 selected users, the hybrid reaches 48.53% test coverage at Top 100, against about 18.8% for the best single embedding. If correct, the result suggests that fusing complementary graph views is a direct route to better social recommendations without using item content or demographics.

What carries the argument

The load-bearing mechanism is a two-stage pipeline. First, the explicit Yelp friend graph is converted into an implicit weighted graph $G'$ by assigning edge weights $W_{ij}=(|L_i\cap L_j|+|D_i\cap D_j|)/(|L_i\cup L_j\cup D_i\cup D_j|)$ over liked and disliked restaurants. Three embeddings are computed on $G'$: spectral embedding from Laplacian eigenvectors, HOPE from a generalized SVD that preserves higher-order proximity, and node2vec from biased random walks. For each of 100 fixed users, the hybrid constructs a feature tensor $X$ of shape $(100,3,1434)$ where each entry records whether each embedding recommended a given restaurant from the union ground-truth set, with labels $Y$ marking the user's own high-rated restaurants; a dense neural network with layers of 32, 64, and 128 neurons, ReLU activations, MSE loss, and the Adam optimizer at learning rate $0.0001$, trained for 40 epochs, learns how to weight the three embeddings' votes.

What would settle it

Run the hybrid with a user-disjoint split: train on 80 of the 100 users and test on the remaining 20, while excluding the test users' high-rated restaurants from the training labels and from the top-user recommendation pool. If coverage on the held-out users falls back to roughly the 18–19% seen with individual embeddings, the claimed hybrid advantage is an artifact of training on the test users.

Watch

Extended reading notes

Core claim

The central discovery claimed is that a recommendation filter trained on the combined outputs of spectral embedding, HOPE, and node2vec outperforms each individual embedding on social-graph-based restaurant recommendation. On a Yelp-derived graph with 14,346 users and 407,495 weighted edges, the individual embeddings all land near 18–19% coverage at Top 100, while the hybrid model reports 48.53% coverage and a MAE of 0.514 on its test set. The paper interprets this as validating the hypothesis that different embedding methods capture complementary aspects of social structure—community, higher-order proximity, and random-walk neighborhoods—so their union carries more signal than any one view.

Load-bearing premise

The evaluation assumes that the same 100 users can be used both to train the hybrid model and to measure its test coverage, so if the model has effectively seen those users' high-rated restaurants during training, the reported 48.53% coverage is not a fair measure of how it would perform for a new user.

Editorial extensions

If this is right

  • A social graph alone, with no item text or user demographics, can support restaurant recommendations: node2vec alone reaches 18.84% coverage and the hybrid reaches 48.53% at Top 100.
  • Combining embeddings that preserve different graph structure—community, higher-order proximity, and random-walk neighborhoods—carries more recommendation signal than any single embedding.
  • A small dense network (32-64-128 neurons, 40 epochs) can learn to weight per-restaurant votes from three embeddings using a training set of just 100 users.
  • The paper reports a large gap between train coverage (63.21%) and test coverage (48.53%), and attributes it to users making individual choices that friends cannot predict, leaving room for better modeling of that residual randomness.

Reading between the lines

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

  • Editorial inference: the reported gain would be more convincing if the hybrid were also compared against a pure neighborhood baseline that ignores embeddings, since the 'nearest ten neighbours weighted vote' step may carry much of the recommendation signal on its own.
  • Editorial inference: a user-disjoint evaluation—train on some users, test on others—would tell whether the 48.53% coverage reflects a generalizable fusion rule or memorization of the 100 fixed users; the paper's own train-test gap suggests this matters.
  • Editorial inference: if the complementary-aspects story is right, adding further embeddings (for example, graph convolutional or heterogeneous user-item embeddings) should push coverage higher, and the same architecture should transfer to movie or music social graphs without content features.
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 studies three graph-embedding techniques (spectral embedding, HOPE, and node2vec) for restaurant recommendation from the Yelp social graph. It constructs a weighted user-user graph, obtains per-embedding restaurant recommendations from each user's nearest neighbors, and combines the three recommendation outputs with a deep neural network trained to predict each user's high-rated restaurants. The hybrid is reported to achieve 48.53% test coverage at Top-100 versus 18.84% for the best individual embedding (node2vec), and the authors conclude that the hybrid model validates their hypothesis that combining embeddings improves recommendation quality. The paper also reports MAE values, discusses future work, and provides a literature review.

Significance. If the reported comparison were trustworthy, the paper would provide simple evidence that fusing complementary graph embeddings improves social recommendation, which is a relevant and timely problem. The paper is clearly written and the problem is well motivated. However, the empirical evaluation is not reproducible as presented: no code or data are provided, the train/test protocol is unspecified, the similarity formula in Section III is malformed, and all results are single-run numbers from a hand-picked set of 100 users. The central empirical claim is therefore not established. The paper contains no machine-checked proofs, reproducible code, or parameter-free derivations that would compensate for these experimental weaknesses.

major comments (4)
  1. [Section IV-C and Table I] The paper never specifies how the 100 selected users are partitioned between training the deep network and computing the Hybrid(test) metric. Since X and Y are constructed for exactly the same 100 users (Section IV-C: 'A set of 100 well connected users are chosen for whom, restaurants are recommended and MAE is calculated'), any overlap between training rows and test rows allows the network to memorize which of the 1434 restaurants each user rated. The reported gap between Hybrid(test) at 48.53% and node2vec at 18.84% is therefore uninterpretable. Section VI's admission of a 'huge gap between train and test results' is consistent with memorization rather than generalization. Please specify a user-disjoint (or at least item-disjoint) split, report per-fold or cross-validated results, and apply the same protocol to all baselines.
  2. [Section III] The similarity weight in Section III is printed as W_ij = |L_i∩L_j| + (|D_i∩D_j| / |L_i∪L_j∪D_i∪D_j|, with an unbalanced parenthesis and no normalization applied to the first intersection term. Since this weight is used to construct the weighted graph G' that all embeddings receive, the formula must be corrected before the experimental setup is reproducible. As written, the formula also appears to give a liked-restaurant overlap and a normalized disliked-restaurant overlap, which is not the 'similarity score' described in the surrounding text.
  3. [Section V, Table I] The central comparison is structurally unfair. The hybrid is trained with supervision on the ground-truth labels Y, while the individual-embedding baselines receive no such training on those labels. The reported outperformance therefore conflates the benefit of fusing embeddings with the benefit of supervised fitting. To support the paper's hypothesis, the baselines need the same access to training labels (for example, a learned weighting of neighbor recommendations), or the hybrid must be evaluated in a label-free setting.
  4. [Section V] All reported numbers are single runs on one hand-picked set of 100 users, with no error bars, no cross-validation, and no significance tests. Given the many manually tuned hyperparameters (embedding dimension D, cluster count, number of nearest neighbors, rating bounds, DNN layer sizes, learning rate, and epoch count), the headline difference of 48.53% versus 18.84% cannot be assessed as statistically reliable. The paper should report variance across multiple user samples or random seeds.
minor comments (5)
  1. [Section III] The text says 'movies' where the domain is restaurants; for example, 'total number of movies both have seen' should read 'restaurants.'
  2. [Section IV-B] The number of clusters k for spectral embedding and the node2vec hyperparameters p and q are not reported, which prevents replication of the baseline results.
  3. [Section V] In the MAE formula, the variables N_r and N_hit are used but the surrounding text defines N as the test-set size; the definitions are inconsistent, and the word 'movies' appears again in a restaurant-recommendation context.
  4. [Table I] The hybrid rows only report Top-100 values; the Top-200 columns are empty, so the comparison against node2vec at Top-200 (35.76% coverage) is incomplete.
  5. [Figures] Figure captions are mismatched: Figure 1 is described as 'Architecture for hybrid Recommendation', Figure 2 as 'Deep learning Network Architecture', and Figure 3 as 'Coverage Graph', but the text in Sections IV-C and V does not clearly map these captions to the discussion.

Circularity Check

1 steps flagged · score 6.0 of 10

Hybrid 'test' coverage is measured against the same ground-truth labels used to train it; the comparison to untrained baselines makes the outperformance a fitted result rather than an independent prediction.

  1. fitted input called prediction [Section IV-C (Hybrid Recommendation), Y-label construction; Section V, Table I]
    "we work on a constant ground by using a few selected users throughout the recommendation process. A set of 100 well connected users are chosen for whom, restaurants are recommended and MAE is calculated. ... In this vector Yi, the ground truth restaurants for ui are represented by 1 and rest by 0. The vector [Y1, Y2, ..Yn] is nothing but the labels Y for our data set, with the shape (100, 1434)."

    The DNN is trained with MSE to map X (embedding recommendations for the 100 chosen users) to Y (the same users' ground-truth high-rated restaurants). Table I then reports 'Hybrid(test) 48.53' as if it were a prediction. Because the manuscript specifies no user-disjoint or item-disjoint split—only a 'constant ground' of the same 100 users for both dataset construction and MAE calculation—the test coverage is measured against the very labels the network was optimized to reproduce. The individual-embedding baselines are never fitted to Y, so the comparison is asymmetric: a fitted model is compared with unfitted baselines, and the claimed 48.53% coverage is partly a report of training fit rather than an independent validation of the hybrid hypothesis.

full rationale

The core empirical claim is that the hybrid DNN outperforms individual embeddings. The only load-bearing step that looks circular is the construction of the hybrid's supervision: Y is the ground-truth high-rated-restaurant vector for the same 100 users whose MAE/coverage is reported, and no split is described. That is pattern 2: the 'prediction' is the output of a network trained on the target labels, while baselines get no label training. This makes the superiority of hybrid in Table I a fitted value by construction, unless an undocumented split exists. No pattern 1, 3, 4, 5, or 6 is present: the embeddings are standard external methods, and the paper does not lean on self-citations or uniqueness theorems. The train-test gap noted in Section VI ('there is a huge gap between train and test results') is consistent with memorization rather than generalization and reinforces the concern. Score 6 because one or more reported 'predictions' reduce to a supervised fit on the evaluation labels; the paper is not wholly tautological, but the central comparison is structurally biased.

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

The central claim rests on the graph-construction choices (weights, thresholds), the embedding hyperparameters, and the evaluation protocol; all are either hand-chosen or fitted to the same data used for evaluation.

free parameters (5)
  • embedding dimension D = 25
    Set to 25 for all three embeddings with the justification that it 'clearly suffices', not chosen by any experiment; affects every downstream recommendation.
  • node2vec cluster count k = 3
    Chosen by an unspecified elbow method on an unspecified quality measure; the recommendation step assigns users to clusters, so k changes the nearest-neighbor sets.
  • nearest neighbors count = 10
    The recommendation set for each user is built from the high-rated restaurants of the 10 nearest users in the predicted cluster; the choice 10 is not justified.
  • high-rated restaurant bounds per user = not specified
    Lower and upper limits are placed on each user's high-rated set to avoid empty or dominating recommendation sets; the actual bounds are not reported.
  • DNN hyperparameters (layer sizes, learning rate, epochs) = 32,64,128; 0.0001; 40
    Chosen by monitoring validation loss and described as 'most optimal to the best of our knowledge' in Section IV-C; the final test numbers depend on them.
assumptions (4)
  • domain assumption Social influence alone drives item consumption, so recommendations from social-graph neighbors are suitable candidates
    The entire recommendation pipeline assumes users pick restaurants their friends like; the paper does not measure the strength of this signal on Yelp beyond the reported results.
  • domain assumption The three embedding methods (spectral, HOPE, node2vec) preserve the social structure needed for nearest-neighbor recommendations
    The method relies on the cited properties of each embedding without validating that cosine or nearest-neighbor in the 25-d space corresponds to shared restaurant preferences.
  • ad hoc to paper The similarity weight W_ij = |L_i∩L_j| + |D_i∩D_j| / |L_i∪L_j∪D_i∪D_j| is a valid measure of social proximity
    This formula is introduced in Section III; as printed it is not normalized for the first term, and the normalization denominator is unclear, so the graph weights may be dominated by the number of common liked restaurants.
  • domain assumption Ground truth is the user's own high-rated restaurants, i.e., high ratings are the correct recommendation targets
    The threshold for 'high' rating is not stated and no held-out ratings are used; a user's historical high-rated restaurants may not reflect future preferences.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Graph Embedding Based Hybrid Social Recommendation System." pith.science (2026). https://pith.science/paper/RNEF7UXT

@misc{pith2026190809454,
  author       = {Pith},
  title        = {Pith review of: Graph Embedding Based Hybrid Social Recommendation System},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RNEF7UXT}},
  note         = {Machine review of arXiv:1908.09454}
}
read the original abstract

Item recommendation tasks are a widely studied topic. Recent developments in deep learning and spectral methods paved a path towards efficient graph embedding techniques. But little research has been done on applying these graph embedding to social graphs for recommendation tasks. This paper focuses at performance of various embedding methods applied on social graphs for the task of item recommendation. Additionally, a hybrid model is proposed wherein chosen embedding models are combined together to give a collective output. We put forward the hypothesis that such a hybrid model would perform better than individual embedding for recommendation task. With recommendation using individual embedding as a baseline, performance for hybrid model for the same task is evaluated and compared. Standard metrics are used for qualitative comparison. It is found that the proposed hybrid model outperforms the baseline.

Figures

Figures reproduced from arXiv: 1908.09454 by the authors.

Figure 1
Figure 1. Our Architecture for hybrid Recommendation [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 4
Figure 4. Mean Average Error Graph for Embedding Recommendation Method Coverage % MAE Top 100 Top 200 Top 100 Top 200 Spectral Clustering 18.62 28.08 0.799 0.685 HOPE 18.34 27.57 0.800 0.689 node2vec 18.84 35.76 0.801 0.630 Hybrid(train) 63.21 0.367 Hybrid(test) 48.53 0.514 TABLE I COMPARISON OF RESULTS FROM DIFFERENT RECOMMENDATION METHODS results are summarized in Table I. The hybrid model entries in the table shown, are fo… view at source ↗
Figure 2
Figure 2. Our Deep learning Network Architecture taken from Tensorboard [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figures from the paper (1 more)
Figure 3
Figure 3. Figure 3: Coverage Graph for Embedding where Nu is the number of movies actually rated by the user U. Coverage estimates the fraction of movies that were rated by user covered by the recommendation system. Before going into the final metric comparisons, we com￾pared each of the …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

38 extracted references · 30 canonical work pages

  1. [1]

    Social network data mining: Research questions, techniques, and applications,

    N. Memon, J. J. Xu, D. L. Hicks, and H. Chen, “Social network data mining: Research questions, techniques, and applications,” in Data mining for social network data , pp. 1–7, Springer, 2010

  2. [2]

    Deep learning based rec- ommender system: A survey and new perspectives,

    S. Zhang, L. Yao, A. Sun, and Y . Tay, “Deep learning based rec- ommender system: A survey and new perspectives,” ACM Computing Surveys (CSUR), vol. 52, no. 1, p. 5, 2019

  3. [3]

    Wasserman and K

    S. Wasserman and K. Faust, Social network analysis: Methods and applications, vol. 8. Cambridge university press, 1994

  4. [4]

    A comprehensive survey of graph embedding: Problems, techniques, and applications,

    H. Cai, V . W. Zheng, and K. C.-C. Chang, “A comprehensive survey of graph embedding: Problems, techniques, and applications,” IEEE Transactions on Knowledge and Data Engineering , vol. 30, no. 9, pp. 1616–1637, 2018

  5. [5]

    Yelp dataset challenge,

    Y . D. Challenge, “Yelp dataset challenge,” 2013

  6. [6]

    A social network-based recommender system (snrs),

    J. He and W. W. Chu, “A social network-based recommender system (snrs),” in Data mining for social network data , pp. 47–74, Springer, 2010

  7. [7]

    User’s interests- based movie recommendation in heterogeneous network,

    W. Yang, X. Cui, J. Liu, Z. Wang, W. Zhu, and L. Wei, “User’s interests- based movie recommendation in heterogeneous network,” in 2015 In- ternational Conference on Identification, Information, and Knowledge in the Internet of Things (IIKI) , pp. 74–77, IEEE, 2015

  8. [8]

    A survey of active learning in collaborative filtering recommender systems,

    M. Elahi, F. Ricci, and N. Rubens, “A survey of active learning in collaborative filtering recommender systems,”Computer Science Review, vol. 20, pp. 29–50, 2016

Show all 38 references
  1. [9]

    Introduction to recommender systems handbook,

    F. Ricci, L. Rokach, and B. Shapira, “Introduction to recommender systems handbook,” in Recommender systems handbook , pp. 1–35, Springer, 2011

  2. [10]

    Recommender systems survey,

    J. Bobadilla, F. Ortega, A. Hernando, and A. Guti ´errez, “Recommender systems survey,”Knowledge-based systems, vol. 46, pp. 109–132, 2013

  3. [11]

    Prof. a. thomas,survey on recommendation system meth- ods,

    P. Nagarnaik, “Prof. a. thomas,survey on recommendation system meth- ods,” in International Conference On Eletronics And Communication System (ICECS), 2015

  4. [12]

    A survey paper on recommender systems,

    D. Almazro, G. Shahatah, L. Albdulkarim, M. Kherees, R. Martinez, and W. Nzoukou, “A survey paper on recommender systems,”arXiv preprint arXiv:1006.5278, 2010

  5. [13]

    Modeling the assimilation-contrast ef- fects in online product rating systems: Debiasing and recommendations,

    X. Zhang, J. Zhao, and J. Lui, “Modeling the assimilation-contrast ef- fects in online product rating systems: Debiasing and recommendations,” in Proceedings of the Eleventh ACM Conference on Recommender Systems, pp. 98–106, ACM, 2017

  6. [14]

    Hybrid recommender systems using social net- work analysis,

    K.-J. Kim and H. Ahn, “Hybrid recommender systems using social net- work analysis,” in Proceedings of World Academy of Science, Engineer- ing and Technology , no. 64, World Academy of Science, Engineering and Technology, 2012

  7. [15]

    Trust-based recommender systems: an overview,

    A. Selmi, Z. Brahmi, and M. M. Gammoudi, “Trust-based recommender systems: an overview,” in Proceedings of 27th International Business Information Management Association (IBIMA) Conference, Milan, Italy, 2016

  8. [16]

    Cross-domain recommendation via clustering on multi-layer graphs,

    A. Farseev, I. Samborskii, A. Filchenkov, and T.-S. Chua, “Cross-domain recommendation via clustering on multi-layer graphs,” in Proceedings of the 40th International ACM SIGIR Conference on Research and Development in Information Retrieval , pp. 195–204, ACM, 2017

  9. [17]

    Representation learning on graphs: Methods and applications,

    W. L. Hamilton, R. Ying, and J. Leskovec, “Representation learning on graphs: Methods and applications,” arXiv preprint arXiv:1709.05584 , 2017

  10. [18]

    Graph embedding techniques, applications, and performance: A survey,

    P. Goyal and E. Ferrara, “Graph embedding techniques, applications, and performance: A survey,”Knowledge-Based Systems, vol. 151, pp. 78–94, 2018

  11. [19]

    Asymmetric transitivity preserving graph embedding,

    M. Ou, P. Cui, J. Pei, Z. Zhang, and W. Zhu, “Asymmetric transitivity preserving graph embedding,” inProceedings of the 22nd ACM SIGKDD international conference on Knowledge discovery and data mining , pp. 1105–1114, ACM, 2016

  12. [20]

    Laplacian eigenmaps and spectral techniques for embedding and clustering,

    M. Belkin and P. Niyogi, “Laplacian eigenmaps and spectral techniques for embedding and clustering,” in Advances in neural information processing systems, pp. 585–591, 2002

  13. [21]

    Distributed large-scale natural graph factorization,

    A. Ahmed, N. Shervashidze, S. Narayanamurthy, V . Josifovski, and A. J. Smola, “Distributed large-scale natural graph factorization,” in Proceedings of the 22nd international conference on World Wide Web , pp. 37–48, ACM, 2013

  14. [22]

    Deepwalk: Online learning of social representations,

    B. Perozzi, R. Al-Rfou, and S. Skiena, “Deepwalk: Online learning of social representations,” in Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining , pp. 701–710, ACM, 2014

  15. [23]

    node2vec: Scalable feature learning for networks,

    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 , pp. 855–864, ACM, 2016

  16. [24]

    Semi-supervised classification with graph convolutional networks,

    T. N. Kipf and M. Welling, “Semi-supervised classification with graph convolutional networks,” arXiv preprint arXiv:1609.02907 , 2016

  17. [25]

    Distributed representations of words and phrases and their composition- ality,

    T. Mikolov, I. Sutskever, K. Chen, G. S. Corrado, and J. Dean, “Distributed representations of words and phrases and their composition- ality,” in Advances in neural information processing systems , pp. 3111– 3119, 2013

  18. [26]

    Efficient estimation of word representations in vector space,

    T. Mikolov, K. Chen, G. Corrado, and J. Dean, “Efficient estimation of word representations in vector space,” arXiv preprint arXiv:1301.3781 , 2013

  19. [27]

    Spectral clustering for link prediction in social networks with positive and negative links,

    P. Symeonidis and N. Mantas, “Spectral clustering for link prediction in social networks with positive and negative links,” Social Network Analysis and Mining , vol. 3, no. 4, pp. 1433–1447, 2013

  20. [28]

    Network embedding based recommendation method in social networks,

    Y . Wen, L. Guo, Z. Chen, and J. Ma, “Network embedding based recommendation method in social networks,” in Companion of the The Web Conference 2018 on The Web Conference 2018 , pp. 11–12, International World Wide Web Conferences Steering Committee, 2018

  21. [29]

    Real-time social recommendation based on graph embedding and temporal context,

    P. Liu, L. Zhang, and J. A. Gulla, “Real-time social recommendation based on graph embedding and temporal context,” International Journal of Human-Computer Studies , vol. 121, pp. 58–72, 2019

  22. [30]

    On spectral clustering: Anal- ysis and an algorithm,

    A. Y . Ng, M. I. Jordan, and Y . Weiss, “On spectral clustering: Anal- ysis and an algorithm,” in Advances in neural information processing systems, pp. 849–856, 2002

  23. [31]

    A tutorial on spectral clustering,

    U. V on Luxburg, “A tutorial on spectral clustering,” Statistics and computing, vol. 17, no. 4, pp. 395–416, 2007

  24. [32]

    On spectral graph embedding: A non-backtracking perspective and graph approximation,

    F. Jiang, L. He, Y . Zheng, E. Zhu, J. Xu, and P. S. Yu, “On spectral graph embedding: A non-backtracking perspective and graph approximation,” in Proceedings of the 2018 SIAM International Conference on Data Mining, pp. 324–332, SIAM, 2018

  25. [33]

    Spectral embedding of graphs,

    B. Luo, R. C. Wilson, and E. R. Hancock, “Spectral embedding of graphs,” Pattern recognition, vol. 36, no. 10, pp. 2213–2230, 2003

  26. [34]

    Hybrid recommender systems: A systematic literature review,

    E. C ¸ ano and M. Morisio, “Hybrid recommender systems: A systematic literature review,” Intelligent Data Analysis , vol. 21, no. 6, pp. 1487– 1524, 2017

  27. [35]

    Deep learning. nature 521 (7553): 436,

    Y . LeCun, Y . Bengio, and G. Hinton, “Deep learning. nature 521 (7553): 436,” Google Scholar, 2015

  28. [36]

    Continuous-time dynamic network embeddings,

    G. H. Nguyen, J. B. Lee, R. A. Rossi, N. K. Ahmed, E. Koh, and S. Kim, “Continuous-time dynamic network embeddings,” in Companion of the The Web Conference 2018 on The Web Conference 2018 , pp. 969–976, International World Wide Web Conferences Steering Committee, 2018

  29. [37]

    Dyngem: Deep embedding method for dynamic graphs,

    P. Goyal, N. Kamra, X. He, and Y . Liu, “Dyngem: Deep embedding method for dynamic graphs,” arXiv preprint arXiv:1805.11273 , 2018

  30. [38]

    Attributed network embedding for learning in a dynamic environment,

    J. Li, H. Dani, X. Hu, J. Tang, Y . Chang, and H. Liu, “Attributed network embedding for learning in a dynamic environment,” in Proceedings of the 2017 ACM on Conference on Information and Knowledge Manage- ment, pp. 387–396, ACM, 2017

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.