Hebbian Graph Embeddings

Shalin Shah; Venkataramana Kini

REVIEW 6 major objections 7 minor 24 references

Hebbian Graph Embeddings

T0 review · 6 major / 7 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a minimal Hebbian update rule—propagating Gaussian-perturbed embeddings along edges with annealed variance—learns node embeddings that beat node2vec, GF, SDNE, HOPE, LE, and VGAE on reconstruction and link prediction…

desk verdict A simple, plausibly useful iterative embedding method that is not currently backed by enough evidence to support its headline claims; the link-prediction protocol needs clarification before any comparison is taken seriously. read the letter →

arxiv 1908.08037 v4 pith:Z5EPUW7N submitted 2019-08-21 cs.LG cs.IRstat.ML

Shalin Shah , Venkataramana Kini This is my paper · ORCID

classification cs.LGcs.IRstat.ML

keywords graphembeddingsHebbianlearninglinkpredictionreconstructionrecommendersystemsmessagepassingvarianceannealingerrorless

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 tries to establish that a single error-free Hebbian update—one line of vector addition per edge—is enough to learn useful node embeddings. Each node starts as random Gaussian noise; iteratively, every node receives the perturbed embedding of each neighbor, scaled by the edge transition probability, while a global variance is annealed from 10 downward. The authors argue that this message-passing procedure, inspired by PageRank and simulated annealing, lets long-distance similarities emerge without any deep network or shared-memory machinery. If the empirical comparisons hold, the method is a strong, embarrassingly parallel alternative to node2vec, GF, SDNE, HOPE, LE, and VGAE for link prediction and reconstruction, and it scales to a 13.1-billion-edge recommender graph. The one caveat in the paper's own tables is that SEAL beats the method on three of the four small link-prediction benchmarks.

What carries the argument

The load-bearing object is the Hebbian update in Equations 3-5 (Algorithm 1): for each directed edge $i\to j$, draw a Gaussian perturbation $\tilde{w}_i \sim \mathcal{N}(w_i, \sigma^2 I)$, then add $\delta_j = \sum_i \tilde{w}_i p_{ij} \eta$ to $w_j$, with a random negative edge contributing the negated vector at probability 0.5. The variance $\sigma^2$ starts at 10 and is divided by 1.1 each iteration, the annealing that the authors take from simulated annealing. This rule carries entire embedding vectors—rather than labels or scalars—across the graph, so repeated iterations propagate information beyond immediate neighbors. It is the mechanism that produces the claimed gains: no loss function, no deep network, and no shared memory among the edge updates, which is what makes the method embarrassingly parallel.

What would settle it

Re-run node2vec, GF, SDNE, HOPE, LE, and VGAE on AstroPh, BlogCatalog, and HepTh at 200 dimensions with identical 10% edge holdouts and 1024-node MAP sampling; if the reported gaps in Tables 1 and 5 shrink to near zero, the paper's central empirical claim fails.

Watch

Extended reading notes

Core claim

The central claim is that the iterative update of Algorithm 1 produces embeddings whose mean average precision on reconstruction and link prediction exceeds that of node2vec, GF, SDNE, HOPE, and LE on AstroPh, BlogCatalog, and HepTh, and whose average precision exceeds VGAE on all four small link-prediction benchmarks (Power, PB, USAir, C.Ele) and SEAL on one of those four. The authors model the embedding of each node, in effect, as a non-convex Gaussian mixture of the embeddings of its neighbors: at each iteration the noisy version of each neighbor's vector is added, weighted by the transition probability, and the variance is reduced by a factor of 1.1 until the graph settles. They also report that the same 10-iteration procedure, implemented on Apache Spark, achieves a hit rate of 30-31% at top-10 recommendations on a held-out user sample in a 200,000-node, 13.1-billion-edge retail item graph. The paper's discovery, on its own terms, is that this minimal errorless associative rule is effective at scale.

Load-bearing premise

The reported advantage rests on MAP and average-precision numbers being directly comparable across tables even though Hebbian embeddings run at 200 dimensions while the baselines are reported at 256, and without evidence that the 10% holdout, the 1024-node MAP sampling, and hyperparameters match the baseline protocols.

Editorial extensions

If this is right

If the comparisons are accurate, a 10-iteration, learning-rate-1.0 Hebbian pass is a cheaper and simpler replacement for node2vec-style random-walk training on reconstruction and link-prediction tasks.
Because the update is just vector addition over edges, any large graph that fits a distributed data structure can be embedded; the authors directly demonstrate this on a 200,000-node, 13.1-billion-edge item graph.
Higher dimensionalities (up to 500 in the tables) monotonically improve MAP, so the method converts extra compute into better embeddings without architectural changes.
The negative-edge sampling step means the method does not merely smooth the graph; it also separates unrelated nodes, a design feature the authors identify as part of the update.

Reading between the lines

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

If the update is viewed as a randomized asynchronous power iteration, the convergence rate and the claimed benefit of annealing likely depend on the spectral gap of the transition matrix; a testable corollary is that graphs with slow mixing need more iterations or slower variance reduction.
The reported monotone MAP gains with dimensionality suggest a relatively smooth embedding landscape; a natural extension is to test the same rule on directed, signed, or temporal graphs where symmetric baselines are known to degrade.
The negative-sample term, fixed at probability 0.5, resembles noise contrastive estimation; ablating it would reveal how much of the separation comes from repulsion rather than from Hebbian attraction.

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.

Desk Editor's Note

read the letter

The actual new thing here is a very simple iterative update: initialize random vectors, propagate noisy neighbor embeddings along edges with transition probabilities, anneal the variance, and push random negative edges. I have not seen this exact rule in the cited literature, and it is plausibly effective and embarrassingly parallel. The large-scale recommender result on a 13.1-billion-edge graph is also a real data point, even if there is no baseline to interpret it against. That part deserves credit.

Where the paper gets shaky is in the evidence for “outperforms state of the art.” The single biggest issue is the link-prediction writeup. Section 3.3 says 10% of edges are held out, but it never says those edges are removed from the graph before Algorithm 1 runs. Algorithm 1 takes the full graph as input. If the held-out edges stay in the graph, Table 5 is reconstruction, not link prediction, and the large gap over the random baseline in Table 9 is exactly what you would expect from seeing the test edges during training. I cannot tell from the text which protocol was actually used. That is a load-bearing ambiguity, not a stylistic gap.

There are smaller problems. The comparisons in Tables 1 and 5 use 200 dimensions for the new method while node2vec, GF, SDNE, HOPE, and LE are listed at 256. There are no error bars, no multiple runs, no released code, and the recommender result has no baseline at all. The “Gaussian mixture model” in Eq. (2) is essentially a restatement of the update rule and does not provide independent theoretical grounding. None of these is fatal by itself, but together they mean the central claim is only partially supported.

My own read: the algorithm is coherent, simple, and likely does what the pseudocode says. The math is not the problem. The problem is that the empirical claims are not yet reproducible or fully specified. If the authors clarify that test edges were removed, release code, add error bars, and give a baseline for the recommender, the paper could be a useful industry-flavored contribution. As it stands, I would not cite the numbers.

This deserves a serious referee, not a desk reject. Send it to review with a clear request for code, protocol clarification, and error bars. The reader’s conditional verdict is about right; the stress-test concern is the right one to put to the authors.

Referee Report

6 major / 7 minor

Summary. The paper proposes "Hebbian Graph Embeddings" (HGE), an iterative, message-passing algorithm in which each node embedding is repeatedly updated by adding noise-perturbed embeddings of its neighbors, scaled by transition probabilities, with an annealing schedule that reduces the variance over iterations. The authors report reconstruction and link-prediction mean average precision (MAP) on AstroPh, BlogCatalog, HepTh and further datasets, compare against node2vec, GF, SDNE, HOPE, LE, VGAE, and SEAL, and describe a large-scale retailer recommender experiment with 200,000 nodes and 13.1 billion edges. The core claim is that this simple, embarrassingly parallel update rule produces embeddings that outperform state-of-the-art graph embedding methods on link prediction and reconstruction while scaling to very large graphs.

Significance. If the empirical claims held under a fully matched protocol, the contribution would be significant: an extremely simple, parameter-light, parallelizable embedding update rule with competitive or better performance than much more complex methods, plus a demonstration on an industrial-scale graph. The paper is also honest about its exploratory character and does not oversell the theory: the Gaussian-mixture framing is presented as a model rather than a proof. However, the paper's load-bearing empirical comparison is currently only partially supported. The key assets are the algorithmic simplicity and the breadth of datasets, but the link-prediction protocol ambiguity and the lack of matched baselines mean that the claimed superiority over existing methods is not yet established.

major comments (6)

[§3.3, Algorithm 1, Eq. (7), Tables 5, 6, 8] The link-prediction protocol is not fully specified: the text says 'We keep 10% of the edges as a held out test set' but never states that these held-out edges are removed from the graph before Algorithm 1 runs. Since Algorithm 1 takes the full graph as input and Eq. (7) propagates embeddings over all edges, the reported MAP on the held-out edges could be measuring reconstruction of seen edges rather than prediction of unseen edges. If the held-out edges are not masked, the comparison to node2vec, GF, SDNE, HOPE, LE, and VGAE is invalidated. The authors must state explicitly that test edges are removed before training and, ideally, give the exact masking step in the pseudocode.
[§3.3, Tables 5, 6, 7] The SOTA comparisons are not matched on protocol. The paper compares 200-dimensional Hebbian embeddings against numbers taken from [2] and [8] at 256 dimensions (node2vec, GF, SDNE, HOPE, LE) or at unspecified dimensions (SEAL, VGAE), with no evidence that the train/test split, the 1024-node sampling for MAP, the graph preprocessing, or the evaluation code are identical. Empirical claims of 'outperforming' state-of-the-art methods require either running the baselines under the same code base or at least justifying that the quantities in Tables 1, 5, and 6 are directly comparable. As written, the gaps in Tables 5 and 6 may be protocol artifacts, and the central claim is therefore only partially supported.
[§3.2, Table 4] The recommender-system result is presented without any baseline: the 31.1% HitRate@10 on a 200,000-node, 13.1-billion-edge graph is not compared to any other embedding method, to a random recommendation, or to the currently deployed system. The metric itself is also only loosely defined (one random item from a user's history is selected and a top-10 list is computed; it is not stated whether the seed item is excluded, how the co-viewing graph is thresholded, or how many users are in the test set). Since the conclusion includes 'applied to generate relevant recommendations for a large retailer', this industrial claim needs at least a minimal baseline and a precise metric definition.
[§3.1, §3.3, Tables 2 and 8] All reported MAP numbers appear to be single-run point estimates. The algorithm has randomized initialization (Eq. (1)), randomized noise (Eq. (6)), and randomized negative-edge sampling, so the absence of variance estimates, number of runs, or standard errors makes it difficult to assess whether the differences between HGE and the baselines are statistically meaningful or even stable across seeds. This is especially relevant for the claims of superiority in Tables 1 and 5, where the reported margins are often smaller than what one would expect from run-to-run noise.
[§2, Eq. (2) and Eqs. (3)-(5)] The Gaussian-mixture formulation is not providing independent theoretical support: Eq. (2) essentially restates the update rule by saying that each neighbor embedding is drawn from a Gaussian centered at the current node embedding, but the algorithm then averages these draws through Eq. (7). This framing does not by itself imply convergence, boundedness, or any optimality property, so it should not be presented as a theoretical grounding. A short analysis of what the update minimizes (if anything) or a discussion of when the iterative averaging converges would be more useful than the mixture interpretation.
[§3.3, Tables 6 and 7] The run-time comparison in Table 7 is not apples-to-apples: the HGE numbers include Apache Spark initialization and resource allocation, while SEAL and VGAE are run in a different environment, and no hardware/software details are provided for either side. The claim that HGE is 'much faster' than SEAL on Power but 'slower' on USAir and C.Ele is a useful observation, but it cannot be interpreted as a scalability result without specifying the cluster configuration and the baseline implementation details.

minor comments (7)

[§2, Algorithm 1] The phrase 'The graph is weighted, asymmetric and undirected' is confusing: a graph can be weighted and directed, or weighted and undirected, but 'asymmetric' typically implies directedness. The authors should clarify whether edges are directed or whether the asymmetry refers only to the transition probabilities p_ij.
[§2, Eqs. (4)-(7)] The negative-edge sampling step (random negative edge, transition probability 0.5) is described in the text but is missing from Algorithm 1. Any reader implementing the algorithm from the pseudocode would omit a component that is likely important for link prediction, so it should be included in the pseudocode or in a separate step.
[§3.3] 'As in [2], we sample 1024 nodes for calculation of the MAP' should specify whether the sampling is done on the training edges only, and whether the same sampled node set is used across all methods. This detail is needed for reproducibility.
[§3.2, Table 4] The text says '10% of the users are held out as the test set' but the hit-rate metric is computed per item, not per user. The relationship between the user-level split and the item-level evaluation should be explained more carefully.
[§3.1] The claim that 'the learning rate does not affect the results in any significant way' is not supported by any table or figure. Either remove this assertion or add a small sensitivity experiment in the supplementary material.
[§4] In the sentence 'Its a useful experiment of error-free (errorless) learning on graphs', there is a missing apostrophe ('It's') and the term 'error-free' should be used consistently with the earlier discussion of errorless learning.
[References] References [23] and [24] are mentioned in the conclusion as possible comparisons but are not used in the experiments; either cite them in the introduction as related work or remove them from the comparison sentence.

Circularity Check

2 steps flagged · score 6.0 of 10

Link-prediction section never removes the held-out 10% of edges from G before Algorithm 1; as written, Eq. (7) trains on the test edges, so Tables 5/8 report reconstruction, and Eq. (2) merely restates the update.

fitted input called prediction [Section 3.3 ('Results on Link Prediction'), Algorithm 1 Eq. (7), Tables 5 and 8]
"Link Prediction is the task of trying to predict a link between two nodes that were not part of the training data... We keep 10% of the edges as a held out test set... Algorithm 1 ... 2: Inputs: Weighted, asymmetric and undirected graph ... wj ← wj + η w˜ipij (7)"

The paper defines link prediction as predicting edges 'not part of the training data' and reconstruction as using the entire graph with no train/test separation. Section 3.3 says only that 10% of edges are kept as a held-out test set; it never states that those edges are removed from G before Algorithm 1 runs. Algorithm 1's input is the full graph, and Eq. (7) updates wj by summing over Adj(i), so any held-out edge left in G contributes to the embedding of both endpoints. Under a literal reading, the MAP and average-precision values on the chosen 10% of edges are computed on training edges: the 'prediction' is equivalent to retrieving the input adjacency, i.e., reconstruction.
self definitional [Section 2, Eq. (2) versus Eqs. (3)-(5) and Algorithm 1 Eq. (7)]
"We model the embedding at a node as a non-convex Gaussian mixture of the embeddings of the connected nodes. If there is an edge from node i to node j, the embedding of node j is modeled as follows: wj ∼ N (wi,σ 2I) (2)"

Eq. (2) is offered as the model, but the algorithm immediately implements it as the update: draw w˜i from N(wi, σ2I) and add η w˜i pij to wj. The Gaussian-mixture statement is therefore a probabilistic paraphrase of the Hebbian update, not an independent constraint from which the update is derived. It adds no falsifiable content beyond the update rule itself. This is circular framing rather than circular evidence, and it is not the load-bearing part of the paper: the empirical comparisons do not rely on Eq. (2) being an independent theorem.

full rationale

Score 6 reflects the load-bearing link-prediction section. As written, Section 3.3 defines a held-out set but never removes it from the graph fed to Algorithm 1, and Eq. (7) sums over exactly those edges. Executed literally, every positive test edge is present during training, so Tables 5, 6, and 8 measure reconstruction of seen adjacency rather than prediction of unseen links; that is the one place where a central claim reduces to its own input. The reconstruction experiments (Tables 1-3) are not circular: they intentionally use the full graph without train/test separation, and the 1024-node sampling follows the protocol of the cited survey, so those numbers remain a valid independent empirical result. The large-retailer hit-rate is an internal demonstration with no external baseline; the same missing-removal ambiguity applies to the 10% held-out users because the co-viewing graph is described before the user split, but this result is not used to argue superiority over prior methods. Eq. (2)'s Gaussian-mixture language does restate the update rule, a mild self-definitional framing, and is recorded as a separate step. There is no self-citation chain and no imported uniqueness theorem, so the circularity is concentrated in the unstated link-prediction protocol rather than in the mathematical derivation itself.

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

The central claim rests on several hand-chosen hyperparameters and on the untested assumption that benchmark results from other papers are directly comparable. No new physical or mathematical entities are introduced.

free parameters (6)

Initial variance sigma^2 = 10
Hand-chosen starting noise scale; annealed every iteration, no sensitivity analysis.
Variance reduction factor tau = 1.1
Hand-chosen annealing rate; no sensitivity analysis.
Number of iterations N = 10
Used for all experiments; no convergence criterion or sensitivity analysis.
Learning rate eta = 1.0
Fixed; authors claim learning rate does not matter but show no supporting experiments.
Negative-edge transition probability = 0.5
Fixed in prose for propagating negative embeddings over random non-edges; omitted from Algorithm 1.
Embedding dimensionality K = 200 (headline); 10-500 explored
Output dimensionality; chosen for comparison against 256-dimensional baselines.

assumptions (4)

ad hoc to paper Iterative linear propagation of noisy embeddings produces stable, informative vectors.
The update is an unnormalized linear recurrence with no divergence analysis or normalization step; the paper assumes 10 iterations yield useful embeddings.
domain assumption Transition probabilities pij = #(i->j)/#i encode semantic relatedness.
Used in Eq. (4) and in building the retail co-view graph; reconstruction and link-prediction quality depends on this choice.
domain assumption Inner product between learned embeddings is a valid similarity score.
Evaluation generates top-k neighbors by inner product, so the vector geometry must match graph proximity.
domain assumption Benchmark numbers from [2] and [8] are protocol-comparable to the authors' runs.
The paper states its results are directly comparable to [2]; any mismatch in splits, sampling, or hyperparameters affects the outperformance claim.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Hebbian Graph Embeddings." pith.science (2026). https://pith.science/paper/Z5EPUW7N

@misc{pith2026190808037,
  author       = {Pith},
  title        = {Pith review of: Hebbian Graph Embeddings},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Z5EPUW7N}},
  note         = {Machine review of arXiv:1908.08037}
}

read the original abstract

Representation learning has recently been successfully used to create vector representations of entities in language learning, recommender systems and in similarity learning. Graph embeddings exploit the locality structure of a graph and generate embeddings for nodes which could be words in a language, products of a retail website; and the nodes are connected based on a context window. In this paper, we consider graph embeddings with an error-free associative learning update rule, which models the embedding vector of node as a non-convex Gaussian mixture of the embeddings of the nodes in its immediate vicinity with some constant variance that is reduced as iterations progress. It is very easy to parallelize our algorithm without any form of shared memory, which makes it possible to use it on very large graphs with a much higher dimensionality of the embeddings. We study the efficacy of proposed method on several benchmark data sets and favorably compare with state of the art methods. Further, proposed method is applied to generate relevant recommendations for a large retailer.

Figures

Figures reproduced from arXiv: 1908.08037 by the authors.

**Figure 2.** Mean Average Precision for Reconstruction with Varying Dimensionality. [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

**Figure 3.** Mean Average Precision for Link Prediction with Varying Dimensionality. [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

24 extracted references · 18 canonical work pages

[2]

Graph embedding techniques, applications, and performance: A survey

Palash Goyal and Emilio Ferrara. Graph embedding techniques, applications, and performance: A survey. Knowledge-Based Systems, 151:78–94, 2018

2018
[8]

Link prediction based on graph neural networks

Muhan Zhang and Yixin Chen. Link prediction based on graph neural networks. In Advances in Neural Information Processing Systems , pages 5165–5175, 2018. 10 Figure 3: Mean Average Precision for Link Prediction with Varying Dimensionality

work page 2018
[1]

A comprehen- sive survey of graph embedding: Problems, techniques, and applications

Hongyun Cai, Vincent W Zheng, and Kevin Chen-Chuan Chang. A comprehen- sive survey of graph embedding: Problems, techniques, and applications. IEEE Transactions on Knowledge and Data Engineering , 30(9):1616–1637, 2018

work page 2018
[3]

node2vec: Scalable feature learning for net- works

Aditya Grover and Jure Leskovec. node2vec: Scalable feature learning for net- works. In Proceedings of the 22nd ACM SIGKDD international conference on Knowledge discovery and data mining , pages 855–864. ACM, 2016. 9 Table 8: Mean Average Precision (MAP) results for network embeddings for Link Prediction (10% randomly chosen edges are held out as the test...

work page 2016
[4]

Dynamic probabilistic models for latent feature propagation in social networks

Creighton Heaukulani and Zoubin Ghahramani. Dynamic probabilistic models for latent feature propagation in social networks. In International Conference on Machine Learning, pages 275–283, 2013

work page 2013
[5]

Spread of information through a population with socio- structural bias: I

Anatol Rapoport. Spread of information through a population with socio- structural bias: I. assumption of transitivity. The bulletin of mathematical bio- physics, 15(4):523–533, 1953

work page 1953
[6]

The pager- ank citation ranking: Bringing order to the web

Lawrence Page, Sergey Brin, Rajeev Motwani, and Terry Winograd. The pager- ank citation ranking: Bringing order to the web. Technical report, Stanford Info- Lab, 1999

work page 1999
[7]

Textrank: Bringing order into text

Rada Mihalcea and Paul Tarau. Textrank: Bringing order into text. InProceedings of the 2004 conference on empirical methods in natural language processing , pages 404–411, 2004

work page 2004

Show all 24 references

[9]

Variational graph auto-encoders

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

2016 arXiv
[10]

Distributed large-scale natural graph factorization

Amr Ahmed, Nino Shervashidze, Shravan Narayanamurthy, Vanja Josifovski, and Alexander J Smola. Distributed large-scale natural graph factorization. In Pro- ceedings of the 22nd international conference on World Wide Web , pages 37–48. ACM, 2013

2013
[11]

Structural deep network embedding

Daixin Wang, Peng Cui, and Wenwu Zhu. Structural deep network embedding. In Proceedings of the 22nd ACM SIGKDD international conference on Knowledge discovery and data mining , pages 1225–1234. ACM, 2016

2016
[12]

Asymmet- ric transitivity preserving graph embedding

Mingdong Ou, Peng Cui, Jian Pei, Ziwei Zhang, and Wenwu Zhu. Asymmet- ric transitivity preserving graph embedding. In Proceedings of the 22nd ACM SIGKDD international conference on Knowledge discovery and data mining , pages 1105–1114. ACM, 2016

2016
[13]

Laplacian eigenmaps and spectral techniques for embedding and clustering

Mikhail Belkin and Partha Niyogi. Laplacian eigenmaps and spectral techniques for embedding and clustering. In Advances in neural information processing systems, pages 585–591, 2002

2002
[14]

Nonlinear dimensionality reduction by locally linear embedding

Sam T Roweis and Lawrence K Saul. Nonlinear dimensionality reduction by locally linear embedding. science, 290(5500):2323–2326, 2000. 11

2000
[15]

Optimization by simu- lated annealing

Scott Kirkpatrick, C Daniel Gelatt, and Mario P Vecchi. Optimization by simu- lated annealing. science, 220(4598):671–680, 1983

1983
[16]

The Organization of Behavior

Donald Olding Hebb. The Organization of Behavior. Wiley & Sons, 1949

1949
[17]

Hebbian learning and predictive mirror neurons for actions, sensations and emotions

Christian Keysers and Valeria Gazzola. Hebbian learning and predictive mirror neurons for actions, sensations and emotions. Philosophical Transactions of the Royal Society B: Biological Sciences , 369(1644):20130175, 2014

2014
[18]

Network-oriented modeling

Jan Treur. Network-oriented modeling. Springer, 2016

2016
[19]

Efﬁcient estimation of word representations in vector space

Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. Efﬁcient estimation of word representations in vector space. arXiv preprint arXiv:1301.3781, 2013

2013 arXiv
[20]

How far can you go with hebbian learning, and when does it lead you astray? Processes of change in brain and cognitive development: Attention and performance xxi , 21:33–69, 2006

James L McClelland. How far can you go with hebbian learning, and when does it lead you astray? Processes of change in brain and cognitive development: Attention and performance xxi , 21:33–69, 2006

2006
[21]

Gem: A python package for graph embedding methods

Palash Goyal and Emilio Ferrara. Gem: A python package for graph embedding methods. J. Open Source Software, 3(29):876, 2018

2018
[22]

Poincar ´e embeddings for learning hier- archical representations

Maximillian Nickel and Douwe Kiela. Poincar ´e embeddings for learning hier- archical representations. In Advances in neural information processing systems , pages 6338–6347, 2017

2017
[23]

Hyper- bolic recommender systems

Tran Dang Quang Vinh, Yi Tay, Shuai Zhang, Gao Cong, and Xiao-Li Li. Hyper- bolic recommender systems. arXiv preprint arXiv:1809.01703, 2018

2018 arXiv
[24]

Scalable hyperbolic recommender sys- tems

Benjamin Paul Chamberlain, Stephen R Hardwick, David R Wardrope, Fabon Dzogang, Fabio Daolio, and Sa´ul Vargas. Scalable hyperbolic recommender sys- tems. arXiv preprint arXiv:1902.08648, 2019. 12

1902 arXiv

Pith tools

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

[1] [2]

Graph embedding techniques, applications, and performance: A survey

Palash Goyal and Emilio Ferrara. Graph embedding techniques, applications, and performance: A survey. Knowledge-Based Systems, 151:78–94, 2018

2018

[2] [8]

Link prediction based on graph neural networks

Muhan Zhang and Yixin Chen. Link prediction based on graph neural networks. In Advances in Neural Information Processing Systems , pages 5165–5175, 2018. 10 Figure 3: Mean Average Precision for Link Prediction with Varying Dimensionality

work page 2018

[3] [1]

A comprehen- sive survey of graph embedding: Problems, techniques, and applications

Hongyun Cai, Vincent W Zheng, and Kevin Chen-Chuan Chang. A comprehen- sive survey of graph embedding: Problems, techniques, and applications. IEEE Transactions on Knowledge and Data Engineering , 30(9):1616–1637, 2018

work page 2018

[4] [3]

node2vec: Scalable feature learning for net- works

Aditya Grover and Jure Leskovec. node2vec: Scalable feature learning for net- works. In Proceedings of the 22nd ACM SIGKDD international conference on Knowledge discovery and data mining , pages 855–864. ACM, 2016. 9 Table 8: Mean Average Precision (MAP) results for network embeddings for Link Prediction (10% randomly chosen edges are held out as the test...

work page 2016

[5] [4]

Dynamic probabilistic models for latent feature propagation in social networks

Creighton Heaukulani and Zoubin Ghahramani. Dynamic probabilistic models for latent feature propagation in social networks. In International Conference on Machine Learning, pages 275–283, 2013

work page 2013

[6] [5]

Spread of information through a population with socio- structural bias: I

Anatol Rapoport. Spread of information through a population with socio- structural bias: I. assumption of transitivity. The bulletin of mathematical bio- physics, 15(4):523–533, 1953

work page 1953

[7] [6]

The pager- ank citation ranking: Bringing order to the web

Lawrence Page, Sergey Brin, Rajeev Motwani, and Terry Winograd. The pager- ank citation ranking: Bringing order to the web. Technical report, Stanford Info- Lab, 1999

work page 1999

[8] [7]

Textrank: Bringing order into text

Rada Mihalcea and Paul Tarau. Textrank: Bringing order into text. InProceedings of the 2004 conference on empirical methods in natural language processing , pages 404–411, 2004

work page 2004