Pith. sign in

REVIEW 4 major objections 3 minor 112 references

Structural Alignment in Link Prediction

T0 review · 4 major / 3 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read This thesis argues that knowledge graph learning and link prediction can be modelled entirely from graph structural features, and presents two systems, TWIG and TWIG-I, that instantiate this claim.

desk verdict A careful, open-sourced PhD thesis proposing a structure-first view of link prediction; the framework is coherent but the local-level claim is undercut by the thesis's own seed-sensitivity result. read the letter →

arxiv 2505.04939 v1 pith:2O6KWNH5 submitted 2025-05-08 cs.LG cs.AI

Jeffrey Seathrún Sardina This is my paper · ORCID
classification cs.LGcs.AI
keywords linkpredictionknowledgegraphsgraphstructurestructuralfeaturesembeddingstransferlearninghyperparameterpreferenceAlignment
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 thesis proposes that knowledge graph learning and link prediction can be understood as structural tasks. It argues that a small set of frequency-based graph features, computed from the training triples alone, is enough to predict both how well an embedding model will perform and which missing links are plausible. To support the claim, the author builds two systems: TWIG, which simulates the output of embedding models from structure plus hyperparameters, and TWIG-I, which performs link prediction directly from structure with no learned embeddings. The reported experiments suggest the structural approach is viable, competitive with embedding-based models, and enables cross-graph transfer learning. The thesis states explicitly that it aims to show feasibility, not optimality.

What carries the argument

The load-bearing object is the 22-dimensional structural feature vector defined in Table 3.1. Six fine-grained features describe the core triple directly: subject degree, object degree, predicate frequency, subject-predicate co-frequency, object-predicate co-frequency, and subject-object co-frequency. Sixteen coarse-grained features summarise the neighbourhoods of the subject and object: min, max, and mean neighbour degree; number of distinct neighbours; min, max, and mean frequency of incident edges; and number of distinct incident edges. All features are computed from the training split only, in a position-aware way, so that directionality is preserved and no test information leaks. This vector turns a triple into a fixed-size, label-free representation; TWIG and TWIG-I are then neural networks that map these representations, together with hyperparameters in TWIG's case, to predicted ranks and plausibility scores.

What would settle it

Take two triples in a knowledge graph that receive identical 22-dimensional structural feature vectors but whose correct completions differ; if an embedding model ranks both correctly while a structure-only predictor cannot, the claim that the chosen structural features are sufficient for link prediction is refuted.

Watch

Extended reading notes

Core claim

Structural Alignment is the claim that the structural features of a training knowledge graph, such as node degrees, relation frequencies, co-frequencies, and neighbourhood summaries, can model hyperparameter preference and link prediction performance, and that link prediction can be performed from those features alone, without learned embeddings. The thesis instantiates this claim in two systems. TWIG takes KG structure plus a hyperparameter configuration and predicts the ranks an embedding model would produce, and experiments on three KGEMs and five KGs report that global performance is highly predictable from structure, with MRR correlations above 0.99 across random seeds. TWIG-I replaces learned embeddings with a fixed 22-dimensional structural feature vector for each candidate triple and learns to score triples directly; the reported results show it is competitive with embedding baselines and that pre-training on KGs from other domains improves its accuracy, making cross-KG transfer viable. The author states the aim is feasibility, not optimality, and concludes that a structure-first view of link prediction is viable and useful.

Load-bearing premise

The load-bearing premise is that the 22 hand-selected structural features, computed only from the training split, capture the information a triple and its neighbourhood carry for link prediction; if multi-hop paths, global topology, or label-derived semantics carry essential signal, the central claim gives way.

Editorial extensions

If this is right

  • Hyperparameter search for embedding models could be shortcut by predicting performance from structure plus hyperparameters, skipping expensive training runs.
  • Structure-based link prediction can act as a strong, faster baseline alongside embedding models, with competitive accuracy on standard benchmark KGs.
  • Cross-KG and cross-domain transfer learning works for link prediction when triples are represented by structural features rather than KG-specific embeddings.
  • KGEM outputs can be simulated from structure, meaning at least part of what embedding models learn is expressible in structural terms.
  • KG construction and curation could be guided by structural learnability, since low-degree and low-frequency elements are the hard cases.

Reading between the lines

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

  • A direct test the thesis leaves implicit: permute node and relation labels while preserving all 22 structural features; a purely structural predictor would give identical predictions, so any accuracy change on the relabelled graph would reveal label-derived signal the features miss.
  • The feature set is deliberately local; extending it with multi-hop path counts, motif frequencies, or global degree distributions is a natural next test and could close cases where co-frequency vectors collapse.
  • Because structural encodings are label-free and graph-agnostic, they are a concrete candidate input representation for graph foundation models aiming at cross-graph transfer; the thesis notes the connection but does not build such a model.
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 / 3 minor

Summary. The thesis proposes Structural Alignment, a hypothesis that knowledge-graph learning and link prediction can be modelled as a function of graph structure without learned node or edge embeddings. It instantiates this idea with 22 structural features (six fine-grained and sixteen coarse-grained, all computed only from the training split), and presents two systems: TWIG, which takes structural features and KGEM hyperparameters and predicts KGEM link-prediction performance both globally (MRR) and locally (per-query ranks), and TWIG-I, a structure-based link predictor intended to show that link prediction can be performed directly from structural features and that cross-KG transfer learning is viable. The thesis reports that KGEM hyperparameter preference and performance are determined by structure, that TWIG can predict KGEM output, and that TWIG-I is competitive with state-of-the-art predictors. All code and data are open-sourced, and the thesis includes a bilingual English/Irish presentation.

Significance. If the main claims hold, the work would be a significant reorientation of link prediction research: it would show that embedding-free, structure-only models can both explain KGEM behaviour and perform link prediction, with the additional practical benefit of cross-KG transferability. The manuscript has real strengths: the MRR stability check across random seeds (Section 4.2.1) is a careful reliability control; the worked examples of structural feature computation are concrete; and the open-sourced code and data (TWIG, TWIG-I, PyPI packages, Figshare data) materially support reproducibility. The intended distinction between structural features and learned embeddings is also clearly drawn. However, the local-level claim of TWIG is undercut by the manuscript's own near-zero cross-seed rank correlation result, and the decisive TWIG-I quantitative results are not present in the supplied text. Both issues are load-bearing for the thesis's central conclusions.

major comments (4)
  1. [Section 4.2.1 and Section 4.1.3] The manuscript reports that per-query ranked lists are nearly uncorrelated across random seeds ("the correlation of ranked lists trained with different random initialisations, but identical hyperparameters, was near 0"), while global MRR is very stable (correlation > 0.99 across runs). This directly contradicts the local half of Claim 1 as instantiated by TWIG. TWIG's data model (Section 4.1.3) pairs each query's structural features with the rank assigned by a single KGEM run, and Section 1.1 states that TWIG simulates KGEM output 'both locally and globally'. Since the random seed is not among TWIG's inputs, a target that is essentially seed-specific noise cannot be a deterministic function of the structural features and hyperparameters TWIG receives. Global MRR stability does not rescue local prediction, because an aggregate over many queries can be stable while individual ranks are dominated by noise. The thesis needs either to train TWIG on seed-averaged ranks and demonstrate that those averages are structurally determined, or to explicitly retract the local-level simulation claim; without one of these, the local half of Claim 1 is not supported by the reported evidence.
  2. [Section 5.2.1 and Section 5.4] The supplied version of the manuscript does not contain the quantitative results of the TWIG-I experiments, nor the transfer-learning results. The abstract and summary assert that structure-based link prediction is 'highly effective compared to state-of-the-art approaches' and that cross-KG transfer learning becomes viable, but Section 5.2.1 and Section 5.4 are presented only as headings or descriptions in the text available for review. These numbers are load-bearing for Claim 2 and for one of the thesis's three stated impacts. The manuscript must include the actual results, or must clearly indicate where in the submitted version they appear, before the claims can be verified.
  3. [Section 3.2.1 and Section 6] The Structural Alignment Hypothesis is stated in general terms ('KG learning and link prediction can be modelled as a function of graph structure'), but the experiments instantiate only a single set of 22 hand-selected, 1-hop, frequency-based features. Multi-hop paths, global topological statistics, and label-derived semantic information are excluded by the feature definition. The conclusion that 'link prediction can be understood and modelled as a structural task' is therefore broader than the evidence supplied. For a feasibility claim this scope restriction may be acceptable, but the thesis should either explicitly formulate the validated hypothesis as restricted to this feature family, or add experiments with richer structural features (e.g., path counts, graphlet counts, or global statistics) to show that the omitted information is not needed.
  4. [Section 4.1.2 and Section 4.3] The hyperparameter grid in Table 4.1 fixes batch size, optimiser, and regulariser, and the case-study uses a single KGEM (ComplEx) on a single KG (UMLS). Sections 4.3 and 4.4 are described as extending this to other KGEMs and KGs, but the generalisation claim in Chapter 6 (that Structural Alignment explains KGEM learning 'in the general case') rests on a small set of three KGEMs and five benchmark KGs. This is not necessarily a flaw for a feasibility study, but the general-case phrasing should be tempered to match the empirical coverage, and the effect of the fixed hyperparameters on the conclusions should be discussed explicitly.
minor comments (3)
  1. [Section 3.2.2] In the worked example for (Gondor, At-War-With, Isengard), the subject-side node degrees are listed as 1 (Osgiliath), 2 (Minas Tirith), 2 (Aragorn), 7 (Rohan), and 5 (Isengard), but the text reports the minimum degree as 2; the correct minimum is 1.
  2. [Section 2.4.2] In the summary of Mohamed et al. (2019), the second category of loss functions is labelled 'Pointwise losses' but the sentence describes pairwise losses; this should read 'Pairwise losses' for consistency with the surrounding text.
  3. [Section 4.2.1] The near-zero ranked-list correlation result is reported through a figure and a sentence; a brief table reporting the mean and median correlation across the four runs would make the severity of the local-level noise easier for readers to assess.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the structure-based models are tested on held-out hyperparameters and KGs, and the reported local-seed instability is an empirical limitation rather than a circular derivation.

full rationale

The paper's central derivation chain is empirical rather than definitional. TWIG is trained to map structural features and hyperparameters to KGEM performance, and its main evidence comes from held-out hyperparameter combinations and unseen KGs (Section 4.3), not from re-reporting its training targets. TWIG-I is evaluated as a link predictor on standard benchmark KGs and compared against external KGEM baselines via PyKEEN, so the central claims are not forced by construction. Structural features are computed from the training split only (Section 3.2.1 and Section 4.1.1), while target ranks and MRR come from KGEM evaluation on validation/test splits; no target information is baked into the feature representation. The self-citation in Chapter 4 ('some of the data and methods contained in this chapter have been published in peer-reviewed venues by the author [80, 82]') is a normal reference to prior work and is not load-bearing, since the thesis provides open code, data, and independent evaluations against external benchmarks. The near-zero cross-seed correlation of local ranked lists reported in Section 4.2.1 is a substantive correctness risk for the local half of Claim 1, because per-query ranks appear to be dominated by random seed rather than by the structural features and hyperparameters TWIG receives. However, this is an empirical failure of determinism, not a circularity: no equation in the paper reduces a reported 'prediction' to the fitted input, and the global MRR signal remains independently evaluated. The hand-selected 22-feature set is an ansatz grounded in prior literature, not a renamed version of the target results. Overall, the derivation chain is self-contained and externally testable, so the circularity burden is minimal.

Assumptions & free parameters 4 free parameters · 5 assumptions · 1 invented entities

The central claim rests on a small set of hand-selected structural features, a hand-specified hyperparameter grid, and an assumption that results on a few KGEMs and benchmark KGs generalise. These are the main costs the reader pays upstream; the open-sourced code and data are the compensating evidence, but the decisive empirical sections were not available in the reviewed excerpt.

free parameters (4)
  • KGEM training epochs = 100
    Fixed for all KGEM runs in the case study; chosen for tractability, and it shapes the MRR and rank data TWIG is trained on.
  • Hyperparameter grid ranges = see Table 4.1
    Hand-selected values for embedding dimension, loss, negative sampler, negatives per positive, learning rate, and regularisation; the grid defines the entire TWIG training set.
  • Structural feature aggregation choice
    The decision to summarise neighbourhoods with min, max, and mean statistics rather than full distributions is manual and affects what TWIG and TWIG-I can observe.
  • TWIG/TWIG-I neural network architecture hyperparameters
    The models' own architecture choices and learned parameters are fitted to the data; they are not specified in the reviewed excerpt, and they drive the reported results.
assumptions (5)
  • domain assumption Knowledge graphs are incomplete and link prediction is evaluated under the Open World Assumption.
    Stated in Section 2.2.1; justifies ranking all possible completions rather than assuming absent triples are false.
  • standard math Rank-based metrics such as MRR and Hits@K are valid and sufficient measures of link prediction performance.
    Section 2.2.2 defines these as the evaluation basis; TWIG targets MRR and ranks, so the structural alignment evidence stands or falls with their validity.
  • domain assumption The four frequency-based features (degree, relation frequency, node-relation co-frequency, node-node co-frequency) are the core structural features relevant to link prediction.
    Section 3.2.1 and Table 2.7 adopt these from prior literature; the whole framework rests on them being representative.
  • ad hoc to paper The fixed 22-feature local representation of a triple and its neighbourhood is sufficient for modelling KGEM performance and link prediction.
    Section 3.2.1 fixes coarse-grained aggregates; this is a modelling choice specific to this thesis and is not proven by prior work.
  • ad hoc to paper Observations on ComplEx, DistMult, TransE and a small set of benchmark KGs generalise to KGEMs and KGs in general.
    The thesis generalises to the general case in Section 1.1, but the selected models and KGs are a small sample.
invented entities (1)
  • Structural Alignment Framework
    purpose: Proposed framework that maps KG structural features to KGEM performance, hyperparameter preference, and link prediction results.
    It is the paper's own construct; its only support is the TWIG and TWIG-I experiments reported in the same thesis, so it has no falsifiable handle outside this paper yet.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Structural Alignment in Link Prediction." pith.science (2026). https://pith.science/paper/2O6KWNH5

@misc{pith2026250504939,
  author       = {Pith},
  title        = {Pith review of: Structural Alignment in Link Prediction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2O6KWNH5}},
  note         = {Machine review of arXiv:2505.04939}
}
read the original abstract

While Knowledge Graphs (KGs) have become increasingly popular across various scientific disciplines for their ability to model and interlink huge quantities of data, essentially all real-world KGs are known to be incomplete. As such, with the growth of KG use has been a concurrent development of machine learning tools designed to predict missing information in KGs, which is referred to as the Link Prediction Task. The majority of state-of-the-art link predictors to date have followed an embedding-based paradigm. In this paradigm, it is assumed that the information content of a KG is best represented by the (individual) vector representations of its nodes and edges, and that therefore node and edge embeddings are particularly well-suited to performing link prediction. This thesis proposes an alternative perspective on the field's approach to link prediction and KG data modelling. Specifically, this work re-analyses KGs and state-of-the-art link predictors from a graph-structure-first perspective that models the information content of a KG in terms of whole triples, rather than individual nodes and edges. Following a literature review and two core sets of experiments, this thesis concludes that a structure-first perspective on KGs and link prediction is both viable and useful for understanding KG learning and for enabling cross-KG transfer learning for the link prediction task. This observation is used to create and propose the Structural Alignment Hypothesis, which postulates that link prediction can be understood and modelled as a structural task. All code and data used for this thesis are open-sourced. This thesis was written bilingually, with the main document in English and an informal extended summary in Irish. An Irish-language translation dictionary of machine learning terms (the Focl\'oir Tr\'achtais) created for this work is open-sourced as well.

Figures

Figures reproduced from arXiv: 2505.04939 by the authors.

Figure 1.1
Figure 1.1. An example knowledge graph featuring information about The Lord of the Rings [92], expressed in graphical format. The graphical format of knowledge graphs has allowed them to naturally represent a large variety of real-world data, including social networks, computer networks, biological networks, linguistic data, climate data, general knowledge, and much more [33, 42, 51, 56, 93, 100]. Recently, larger and larger KG… view at source ↗
Figure 1.2
Figure 1.2. An example of knowledge graph embedding based on the KG in [PITH_FULL_IMAGE:figures/full_fig_p024_1_2.png] view at source ↗
Figure 1.3
Figure 1.3. An outline of the Structural Alignment Framework. The Framework begins by taking a set of graph structural features out of the set of all possible graph structural features. It then uses those structural features to directly model KG learning, such as through KGEM simulation or structural link prediction. as a function of KG structural features, and TWIG-I (in Chapter 5), for performing link prediction as a function… view at source ↗
Figures from the paper (27 more)
Figure 2.1
Figure 2.1. Figure 2.1: A visualisation of KGs in the LOD Cloud and the sources from which they were extracted, reproduced with permission from the Linked Open Data Cloud project [20] [PITH_FULL_IMAGE:figures/full_fig_p031_2_1.png]
Figure 2.2
Figure 2.2. Figure 2.2: A visualisation of biological KGs in the LOD Cloud, reproduced with permission from the Linked Open Data Cloud project [20]. A summary of KGs commonly used in the literature, and their general information content domain, is given in [PITH_FULL_IMAGE:figures/full_fig…
Figure 2.3
Figure 2.3. Figure 2.3: An overview of the process of link prediction in knowledge graphs. In essentially all scenarios in which link prediction is applied, there are a plurality of possible answers, not just one. In the current example, multiple completions, includ￾ing (Aragorn, Enemy-Of ,…
Figure 2.4
Figure 2.4. Figure 2.4: An overview of the process of training and evaluating a link predictor using rank-based evaluation [PITH_FULL_IMAGE:figures/full_fig_p037_2_4.png]
Figure 2.5
Figure 2.5. Figure 2.5: An example test set, shown as a graph that could be constructed for the example KG in [PITH_FULL_IMAGE:figures/full_fig_p038_2_5.png]
Figure 2.6
Figure 2.6. Figure 2.6: An overview of the hyperparameters used in KGEMs [PITH_FULL_IMAGE:figures/full_fig_p045_2_6.png]
Figure 2.7
Figure 2.7. Figure 2.7: An overview of the KGEM training loop [PITH_FULL_IMAGE:figures/full_fig_p053_2_7.png]
Figure 2.8
Figure 2.8. Figure 2.8: An overview of the process of hyperparameter selection for KGEMs. In KGEMs, as in machine learning in general, there are several methods by which hyper￾parameter search can be performed: • a grid search, in which all combinations of all hyperparameter values are sear…
Figure 2.9
Figure 2.9. Figure 2.9: An overview of hyperparameter and graph structure influence on KGEM perfor￾mance on the link prediction task, presented as a dependency graph annotated by evidence in the literature. A directed edge from a subject node to an object node indicates that the item subjec…
Figure 3.1
Figure 3.1. Figure 3.1: The regions from which frequency-based structural features are calculated. The core triple is shown in blue; neighbouring triples whose structure also contributes to annotating the core triple are in red; all other triples (whose structures are not used to annotate t…
Figure 3.2
Figure 3.2. Figure 3.2: The regions of the training set from which the structural features for the triple (Aragorn, Enemy-Of , Sauron) in the testing set are calculated. Note that as Aragorn and Suaron are not connected in the training set, these features are necessarily calculated from dis…
Figure 4.1
Figure 4.1. Figure 4.1: An overview of how TWIG learns to simulate KGEMs vs how KGEMs learn to perform link prediction. To do this, the hyperparameters identified in Section 2.3.1 as critical to link prediction were used to construct a hyperparameter grid containing a total of 1215 possible…
Figure 4.2
Figure 4.2. Figure 4.2: A pictorial overview of the TWIG data model. 4.2 Case Study: ComplEx and UMLS In order to assess the viability of a model such as TWIG, a feasibility study was first done on the specific case of the KGEM ComplEx with the KG UMLS. The KGEM ComplEx was used for this in…
Figure 4.3
Figure 4.3. Figure 4.3: Distribution of the correlation values between all ranked lists using the same hyperparameter configurations. This lack of clear correlation suggests that ranked values assigned to individual link prediction queries cannot be used directly as a source for signal in l…
Figure 4.4
Figure 4.4. Figure 4.4: The literature describing hyperparameter and structural dependencies in link prediction (left) and how this corresponds to TWIG’s neural architecture (right). The same colour is used to annotate both the neural network component in TWIG and the literature evidence pr…
Figure 4.5
Figure 4.5. Figure 4.5: A pictorial overview of TWIG’s batch training and evaluation, in which it is trained and evaluated in the context of an entire set of link prediction queries rather than in terms of isolated (single) link prediction queries. parameter combinations it had never seen b…
Figure 4.6
Figure 4.6. Figure 4.6: The 2-phase training method used for TWIG. In order to test TWIG’s ability to simulate the output of ComplEx on UMLS, a simple run of 15 total epochs was performed on the ComplEx-UMLS data using both 1-phase and 2-phase training. The choice of 15 epochs was made beca…
Figure 4.7
Figure 4.7. Figure 4.7: Details of TWIG on ComplEx. The first column shows scatter plots of all ground-truth MRR values vs TWIG’s predicted MRR values for all hyperparameter combi￾nations in the hold-out test set. The second shows the distribution of the ground truth MRR values, and the thi…
Figure 4.8
Figure 4.8. Figure 4.8: Details of TWIG on DistMult. The first column shows scatter plots of all ground-truth MRR values vs TWIG’s predicted MRR values for all hyperparameter combi￾nations in the hold-out test set. The second shows the distribution of the ground truth MRR values, and the th…
Figure 4.9
Figure 4.9. Figure 4.9: Details of TWIG on TransE. The first column shows scatter plots of all ground￾truth MRR values vs TWIG’s predicted MRR values for all hyperparameter combinations in the hold-out test set. The second shows the distribution of the ground truth MRR values, and the third…
Figure 4.10
Figure 4.10. Figure 4.10: A graphical overview of the results of structural and hyperparameter analysis of KGEMs. An edge from a subject node to an object indicates that the subject influences the object. Edge colours and styles denote paths: for example, the solid green edge from “Struc￾tur…
Figure 5.1
Figure 5.1. Figure 5.1: Depiction of the TWIG-I link prediction model. Dataset #Nodes #Predicates #Triples Reference FB15k-237 14505 237 310079 Toutanova et al. [93] WN18RR 40559 11 92583 Toutanova et al. [93] CoDExSmall 2034 42 36543 Safavi et al. [75] DBpedia50 24624 351 34421 Shi et al. …
Figure 5.2
Figure 5.2. Figure 5.2: Visualisations of the structure of subgraphs of FB15k-237, WN18RR, CoDExS￾mall, and DBpedia50. For FB15k-237 and CoDExSmall, 2-hop subgrpahs around an arbitrary node of median degree are shown. For WN18RR and DBpedia50, 6-hop subgrpahs around an arbitrary node of med…
Figure 6.1
Figure 6.1. Figure 6.1: The Link Prediction Pyramid. Under Structural Alignment, the simplest (and most commonly learned) aspect of a KG for link prediction is structure. The next layer up above structure is semantics (and / or ontology) of a KG. Finally, space is left to indicate that othe…
Figure 6.2
Figure 6.2. Figure 6.2: A side-by-side presentation of the Link Prediction Pyramid (left) and the DIKW Pyramid (right). Corresponding layers at higher levels represent abstraction of the layer below and the addition of further, higher-order concepts. information which is non-structural in n…
Figure 6.3
Figure 6.3. Figure 6.3: A graphical expression of the logical rule given in Equation 6.1. Edges in blue represent the antecedent, and the edge in red represents the consequent. Whether Structural Alignment could be extended to elements of logical structure, as well as the implication of suc…
Figure 6.4
Figure 6.4. Figure 6.4: An example of two knowledge graphs, based on the example in [PITH_FULL_IMAGE:figures/full_fig_p155_6_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

112 extracted references · 67 canonical work pages

  1. [1]

    Optuna: A next-generation hyperparameter optimization framework

    Akiba, Takuya, Sano, Shotaro, Yanase, Toshihiko, Ohta, Takeru, and Koyama, Masa- nori. “Optuna: A next-generation hyperparameter optimization framework”. In: Pro- ceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining . 2019, pp. 2623–2631

    2019

  2. [2]

    Bringing Light Into the Dark: A Large-Scale Evaluation of Knowledge Graph Embedding Models Under a Unified Framework

    Ali, Mehdi, Berrendorf, Max, Hoyt, Charles Tapley, Vermue, Laurent, Galkin, Mikhail, Sharifzadeh, Sahand, Fischer, Asja, Tresp, Volker, and Lehmann, Jens. “Bringing Light Into the Dark: A Large-Scale Evaluation of Knowledge Graph Embedding Models Under a Unified Framework”. In:IEEE Transactions on Pattern Analysis and Machine Intelligence 44.12 (2022), pp...

    arXiv 2022

  3. [3]

    PyKEEN 1.0: A Python Library for Training and Evaluating Knowledge Graph Embeddings

    Ali, Mehdi, Berrendorf, Max, Hoyt, Charles Tapley, Vermue, Laurent, Sharifzadeh, Sahand, Tresp, Volker, and Lehmann, Jens. “PyKEEN 1.0: A Python Library for Training and Evaluating Knowledge Graph Embeddings”. In: Journal of Machine Learning Research 22.82 (2021), pp. 1–6. url: http://jmlr.org/papers/v22/20- 825.html

    2021

  4. [4]

    Neural methods for logical reasoning over knowledge gra-phs

    Amayuelas, Alfonso, Zhang, Shuai, Rao, Xi Susie, and Zhang, Ce. “Neural methods for logical reasoning over knowledge gra-phs”. In: International Conference on Learning Representations. 2022

    2022

  5. [5]

    Gephi: an open source software for exploring and manipulating networks

    Bastian, Mathieu, Heymann, Sebastien, and Jacomy, Mathieu. “Gephi: an open source software for exploring and manipulating networks”. In:Proceedings of the international AAAI conference on web and social media . Vol. 3. 1. 2009, pp. 361–362

    2009

  6. [6]

    Bio2RDF: +Towards A Mashup To Build Bioinformatics Knowl- edge System

    Belleau, Fran¸ cois, Nolin, Marc-Alexandre, Tourigny, Nicole, Rigault, Philippe, and Morissette, Jean. “Bio2RDF: +Towards A Mashup To Build Bioinformatics Knowl- edge System”. In: Journal of biomedical informatics 41 (Apr. 2008), pp. 706–16. doi: 10.1016/j.jbi.2008.03.004

    work page doi:10.1016/j.jbi.2008.03.004 2008

  7. [7]

    Adver- sarial Attacks on Knowledge Graph Embeddings via Instance Attribution Methods

    Bhardwaj, Peru, Kelleher, John, Costabello, Luca, and O’Sullivan, Declan. “Adver- sarial Attacks on Knowledge Graph Embeddings via Instance Attribution Methods”. In: Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing. 2021, pp. 8225–8239

    2021

  8. [8]

    Poisoning Knowledge Graph Embeddings via Relation Inference Patterns

    Bhardwaj, Peru, Kelleher, John, Costabello, Luca, and O’Sullivan, Declan. “Poisoning Knowledge Graph Embeddings via Relation Inference Patterns”. In: Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021, pp. 1875–1888...

    2021

Show all 112 references
  1. [9]

    Impli- cations of topological imbalance for representation learning on biomedical knowledge graphs

    Bonner, Stephen, Kirik, Ufuk, Engkvist, Ola, Tang, Jian, and Barrett, Ian P. “Impli- cations of topological imbalance for representation learning on biomedical knowledge graphs”. In: Briefings in Bioinformatics 23.5 (July 2022). bbac279. issn: 1477-4054. doi: 10.1093/bib/bbac2...

    2022 doi

  2. [10]

    A semantic matching energy function for learning with multi-relational data: Application to word- sense disambiguation

    Bordes, Antoine, Glorot, Xavier, Weston, Jason, and Bengio, Yoshua. “A semantic matching energy function for learning with multi-relational data: Application to word- sense disambiguation”. In: Machine learning 94 (2014), pp. 233–259

    2014

  3. [11]

    Translating embeddings for modeling multi-relational data

    Bordes, Antoine, Usunier, Nicolas, Garcia-Duran, Alberto, Weston, Jason, and Yakh- nenko, Oksana. “Translating embeddings for modeling multi-relational data”. In: Ad- vances in neural information processing systems 26 (2013)

    2013

  4. [12]

    Torchkge: Knowledge graph embedding in python and pytorch

    Boschin, Armand. “Torchkge: Knowledge graph embedding in python and pytorch”. In: arXiv preprint arXiv:2009.02963 (2020)

    2020 arXiv

  5. [13]

    On Approximate Reason- ing Capabilities of Low-Rank Vector Spaces

    Bouchard, Guillaume, Singh, Sameer, and Trouillon, Th´ eo. “On Approximate Reason- ing Capabilities of Low-Rank Vector Spaces”. In: AAAI Spring Symposia . 2015. url: https://api.semanticscholar.org/CorpusID:13955854

    2015

  6. [14]

    OpenBioLink: a benchmarking framework for large-scale biomedical lin prkediction

    Breit, Anna, Ott, Simon, Agibetov, Asan, and Samwald, Matthias. “OpenBioLink: a benchmarking framework for large-scale biomedical lin prkediction”. In: Bioinformat- ics 36.13 (2020), pp. 4097–4098

    2020

  7. [15]

    DL-Learner—A frame- work for inductive learning on the Semantic Web

    B¨ uhmann, Lorenz, Lehmann, Jens, and Westphal, Patrick. “DL-Learner—A frame- work for inductive learning on the Semantic Web”. In: Journal of Web Semantics 39 (2016), pp. 15–24

    2016

  8. [17]

    Evaluation of knowledge graph embedding approaches for drug- drug interaction prediction in realistic settings

    Celebi, Remzi, Uyar, Huseyin, Yasar, Erkan, Gumus, Ozgur, Dikenelli, Oguz, and Dumontier, Michel. “Evaluation of knowledge graph embedding approaches for drug- drug interaction prediction in realistic settings”. In: BMC bioinformatics 20 (2019), pp. 1–14

    2019

  9. [18]

    Building a knowledge graph to enable precision medicine

    Chandak, Payal, Huang, Kexin, and Zitnik, Marinka. “Building a knowledge graph to enable precision medicine”. In: Scientific Data 10.1 (2023), p. 67. REFERENCES 215

    2023

  10. [19]

    Hyperparameter search in machine learning

    Claesen, Marc and De Moor, Bart. “Hyperparameter search in machine learning”. In: arXiv preprint arXiv:1502.02127 (2015)

    2015 arXiv

  11. [20]

    The Linked Open Data Cloud

    Cloud, The LoD. The Linked Open Data Cloud . url: https : / / lod - cloud . net/ (visited on 09/23/2024)

    2024

  12. [21]

    AmpliGraph: a library for representation learning on knowledge graphs

    Costabello, Luca, Pai, Sumit, Le Van, Chan, McGrath, Rory, McCarthy, Nicholas, and Tabacof, Pedro. “AmpliGraph: a library for representation learning on knowledge graphs”. In: Retrieved Oct 10 (2019), p. 2019

    2019

  13. [22]

    Towards Understanding the Impact of Graph Structure on Knowledge Graph Embeddings

    Dave, Brandon, Christou, Antrea, and Shimizu, Cogan. “Towards Understanding the Impact of Graph Structure on Knowledge Graph Embeddings”. In: International Con- ference on Neural-Symbolic Learning and Reasoning. Springer. 2024, pp. 41–50

    2024

  14. [23]

    The compara- tive toxicogenomics database: update 2019

    Davis, Allan Peter, Grondin, Cynthia J, Johnson, Robin J, Sciaky, Daniela, McMorran, Roy, Wiegers, Jolene, Wiegers, Thomas C, and Mattingly, Carolyn J. “The compara- tive toxicogenomics database: update 2019”. In: Nucleic acids research 47.D1 (2019), pp. D948–D954

    2019

  15. [24]

    Con- volutional 2d knowledge graph embeddings

    Dettmers, Tim, Minervini, Pasquale, Stenetorp, Pontus, and Riedel, Sebastian. “Con- volutional 2d knowledge graph embeddings”. In: Proceedings of the AAAI conference on artificial intelligence . Vol. 32. 1. 2018

    2018

  16. [25]

    Centrality Measures in multi-layer Knowledge Graphs

    D¨ orpinghaus, Jens, Weil, Vera, D¨ uing, Carsten, and Sommer, Martin W. “Centrality Measures in multi-layer Knowledge Graphs”. In: Annals of Computer Science and Information Systems (2022)

    2022

  17. [26]

    BOHB: Robust and efficient hyper- parameter optimization at scale

    Falkner, Stefan, Klein, Aaron, and Hutter, Frank. “BOHB: Robust and efficient hyper- parameter optimization at scale”. In: International conference on machine learning . PMLR. 2018, pp. 1437–1446

    2018

  18. [27]

    Weighted Random Search for Hyperpa- rameter Optimization

    Florea, Adrian-Catalin and Andonie, Razvan. “Weighted Random Search for Hyperpa- rameter Optimization”. In: INTERNATIONAL JOURNAL OF COMPUTERS COM- MUNICATIONS & CONTROL 14.2 (2019), pp. 154–169

    2019

  19. [28]

    Fast rule mining in ontological knowledge bases with AMIE+

    Gal´ arraga, Luis, Teflioudi, Christina, Hose, Katja, and Suchanek, Fabian M. “Fast rule mining in ontological knowledge bases with AMIE+”. In: The VLDB Journal 24.6 (2015), pp. 707–730

    2015

  20. [29]

    Quantum embedding of knowledge for reason- ing

    Garg, Dinesh, Ikbal, Shajith, Srivastava, Santosh K, Vishwakarma, Harit, Karanam, Hima, and Subramaniam, L Venkata. “Quantum embedding of knowledge for reason- ing”. In: Advances in Neural Information Processing Systems 32 (2019). 216 REFERENCES

    2019

  21. [30]

    Predicting gene disease as- sociations with knowledge graph embeddings for diseases with curtailed information

    Gualdi, Francesco, Oliva, Baldomero, and Pi˜ nero, Janet. “Predicting gene disease as- sociations with knowledge graph embeddings for diseases with curtailed information”. In: NAR Genomics and Bioinformatics 6.2 (2024), lqae049

    2024

  22. [31]

    Elec- tronic health records data summarization for graph machine learning

    Gu´ eret, Christophe Dominique Marie, Costabello, Luca, and Sardina, Jeffrey. Elec- tronic health records data summarization for graph machine learning . US Patent App. 18/333,875. Dec. 2024

    2024

  23. [32]

    Systematic integration of biomedical knowledge prioritizes drugs for repurposing

    Himmelstein, Daniel Scott, Lizee, Antoine, Hessler, Christine, Brueggeman, Leo, Chen, Sabrina L, Hadley, Dexter, Green, Ari, Khankhanian, Pouya, and Baranzini, Sergio E. “Systematic integration of biomedical knowledge prioritizes drugs for repurposing”. In: Elife 6 (2017), e26726

    2017

  24. [33]

    Knowledge Graphs

    Hogan, Aidan, Blomqvist, Eva, Cochez, Michael, D’amato, Claudia, Melo, Gerard De, Gutierrez, Claudio, Kirrane, Sabrina, Gayo, Jos´ e Emilio Labra, Navigli, Roberto, Neumaier, Sebastian, Ngomo, Axel-Cyrille Ngonga, Polleres, Axel, Rashid, Sabbir M., Rula, Anisa, Schmelzeisen, L...

    2021

  25. [34]

    Pygraft: Configurable generation of synthetic schemas and knowledge graphs at your fingertips

    Hubert, Nicolas, Monnin, Pierre, d’Aquin, Mathieu, Monticolo, Davy, and Brun, Armelle. “Pygraft: Configurable generation of synthetic schemas and knowledge graphs at your fingertips”. In: European Semantic Web Conference. Springer. 2024, pp. 3–20

    2024

  26. [35]

    Im- proving knowledge graph embeddings with ontological reasoning

    Jain, Nitisha, Tran, Trung-Kien, Gad-Elrab, Mohamed H, and Stepanova, Daria. “Im- proving knowledge graph embeddings with ontological reasoning”. In: International Semantic Web Conference. Springer. 2021, pp. 410–426

    2021

  27. [36]

    Knowledge base comple- tion: Baseline strikes back (again)

    Jain, Prachi, Rathi, Sushant, Chakrabarti, Soumen, et al. “Knowledge base comple- tion: Baseline strikes back (again)”. In: arXiv preprint arXiv:2005.00804 (2020)

    2020 arXiv

  28. [37]

    Co-embedding of nodes and edges with graph neural networks

    Jiang, Xiaodong, Zhu, Ronghang, Li, Sheng, and Ji, Pengsheng. “Co-embedding of nodes and edges with graph neural networks”. In: IEEE Transactions on Pattern Analysis and Machine Intelligence (2020)

    2020

  29. [38]

    Knowledge Base Completion: Baselines Strike Back

    Kadlec, Rudolf, Bajgar, Ondrej, and Kleindienst, Jan. “Knowledge Base Completion: Baselines Strike Back”. In: ACL 2017 (2017), p. 69

    2017

  30. [39]

    KEGG: Kyoto Encyclopedia of Genes and Ge-nomes

    Kanehisa, Minoru and Goto, Susumu. “KEGG: Kyoto Encyclopedia of Genes and Ge-nomes”. In: Nucleic acids research 28.1 (2000), pp. 27–30

    2000

  31. [40]

    Extending siena to support more expressive and flexible subscriptions

    Keeney, John, Roblek, Dominik, Jones, Dominic, Lewis, David, and O’Sullivan, De- clan. “Extending siena to support more expressive and flexible subscriptions”. In: REFERENCES 217 Proceedings of the second international conference on Distributed event-based systems. 2008, pp. 35–46

    2008

  32. [41]

    Knowledge graphs: A practical review of the research landscape

    Kejriwal, Mayank. “Knowledge graphs: A practical review of the research landscape”. In: Information 13.4 (2022), p. 161

    2022

  33. [42]

    Learning systems of concepts with an infinite relational model

    Kemp, Charles, Tenenbaum, Joshua B, Griffiths, Thomas L, Yamada, Takeshi, and Ueda, Naonori. “Learning systems of concepts with an infinite relational model”. In: AAAI. Vol. 3. 2006, p. 5

    2006

  34. [43]

    Adam: A method for stochastic optimization

    Kingma, Diederik P and Ba, Jimmy. “Adam: A method for stochastic optimization”. In: arXiv preprint arXiv:1412.6980 (2014)

    2014 arXiv

  35. [44]

    Variational graph auto-encoders

    Kipf, Thomas N and Welling, Max. “Variational graph auto-encoders”. In: arXiv pre- print arXiv:1611.07308 (2016)

    2016 arXiv

  36. [45]

    Analysis of the Impact of Negative Sampling on Link Prediction in Knowledge Graphs

    Kotnis, Bhushan and Nastase, Vivi. “Analysis of the Impact of Negative Sampling on Link Prediction in Knowledge Graphs”. In: (Aug. 2017)

    2017

  37. [46]

    EL Embed- dings: Geometric construction of models for the description logic EL++

    Kulmanov, Maxat, Liu-Wei, Wang, Yan, Yuan, and Hoehndorf, Robert. “EL Embed- dings: Geometric construction of models for the description logic EL++”. In: Proceed- ings of the Twenty-Eighth International Joint Conference on Artificial Intelligence. In- ternational Joint Confere...

    2019

  38. [47]

    Canonical tensor de- composition for knowledge base completion

    Lacroix, Timoth´ ee, Usunier, Nicolas, and Obozinski, Guillaume. “Canonical tensor de- composition for knowledge base completion”. In:International Conference on Machine Learning. PMLR. 2018, pp. 2863–2872

    2018

  39. [48]

    Knowledge Graphs Can be Learned with Just Intersection Fea- tures

    Le, Duy, Zhong, Shaochen, Liu, Zirui, Xu, Shuai, Chaudhary, Vipin, Zhou, Kaixiong, and Xu, Zhaozhuo. “Knowledge Graphs Can be Learned with Just Intersection Fea- tures”. In: Forty-first International Conference on Machine Learning

  40. [49]

    A survey of the link prediction on static and temporal knowledge graph

    Le, Thanh, Nguyen, Hoang, and Le, Bac. “A survey of the link prediction on static and temporal knowledge graph”. In: Journal of Research and Development on Information and Communication Technology 2021.2 (2021), pp. 1–34

    2021

  41. [50]

    Pytorch-biggraph: A large scale graph embedding system

    Lerer, Adam, Wu, Ledell, Shen, Jiajun, Lacroix, Timothee, Wehrstedt, Luca, Bose, Abhijit, and Peysakhovich, Alex. “Pytorch-biggraph: A large scale graph embedding system”. In: Proceedings of Machine Learning and Systems 1 (2019), pp. 120–131

    2019

  42. [51]

    Towards a managed extensible control plane for knowledge-based networking

    Lewis, David, Keeney, John, O’Sullivan, Declan, and Guo, Song. “Towards a managed extensible control plane for knowledge-based networking”. In: Large Scale Manage- ment of Distributed Systems: 17th IFIP/IEEE International Workshop on Distributed 218 REFERENCES Systems: Operati...

    2006

  43. [52]

    Towards graph foundation models: A survey and beyond

    Liu, Jiawei, Yang, Cheng, Lu, Zhiyuan, Chen, Junze, Li, Yibo, Zhang, Mengmei, Bai, Ting, Fang, Yuan, Sun, Lichao, Yu, Philip S, et al. “Towards graph foundation models: A survey and beyond”. In: arXiv preprint arXiv:2310.11829 (2023)

    2023 arXiv

  44. [53]

    Neural multi-hop reasoning with logical rules on biomedical knowledge graphs

    Liu, Yushan, Hildebrandt, Marcel, Joblin, Mitchell, Ringsquandl, Martin, Raissouni, Rime, and Tresp, Volker. “Neural multi-hop reasoning with logical rules on biomedical knowledge graphs”. In: The Semantic Web: 18th International Conference, ESWC 2021, Virtual Event, June 6–10...

    2021

  45. [54]

    A knowledge base from multilingual Wikipedias–yago3

    Mahdisoltani, Farzane, Biega, Joanna, and Suchanek, Fabian M. “A knowledge base from multilingual Wikipedias–yago3”. In: Technical report, Telecom ParisTech(2014)

    2014

  46. [55]

    Yago3: A knowl- edge base from multilingual wikipedias

    Mahdisoltani, Farzaneh, Biega, Joanna, and Suchanek, Fabian M. “Yago3: A knowl- edge base from multilingual wikipedias”. In: CIDR. 2013

    2013

  47. [56]

    Aggregating UMLS se- mantic types for reducing conceptual complexity

    McCray, Alexa T, Burgun, Anita, and Bodenreider, Olivier. “Aggregating UMLS se- mantic types for reducing conceptual complexity”. In: Studies in health technology and informatics 84.0 1 (2001), p. 216

    2001

  48. [57]

    Anytime Bottom-Up Rule Learning for Knowledge Graph Com- pletion

    Meilicke, Christian, Chekol, Melisachew Wudage, Ruffinelli, Daniel, and Stucken- schmidt, Heiner. “Anytime Bottom-Up Rule Learning for Knowledge Graph Com- pletion”. In: Proceedings of the Twenty-Eighth International Joint Conference on Ar- tificial Intelligence . Internationa...

    2019

  49. [58]

    Pop- ularity agnostic evaluation of knowledge graph embeddings

    Mohamed, Aisha, Parambath, Shameem, Kaoudi, Zoi, and Aboulnaga, Ashraf. “Pop- ularity agnostic evaluation of knowledge graph embeddings”. In: Conference on Un- certainty in Artificial Intelligence (UAI) . PMLR. 2020, pp. 1059–1068

    2020

  50. [59]

    Biological applications of knowledge graph embedding models

    Mohamed, Sameh K, Nounu, Aayah, and Nov´ aˇ cek, V´ ıt. “Biological applications of knowledge graph embedding models”. In: Briefings in bioinformatics 22.2 (2021), pp. 1679–1693

    2021

  51. [60]

    Discovering protein drug targets using knowledge graph embeddings

    Mohamed, Sameh K, Nov´ aˇ cek, V´ ıt, and Nounu, Aayah. “Discovering protein drug targets using knowledge graph embeddings”. In: Bioinformatics 36.2 (2020), pp. 603– 610

    2020

  52. [61]

    Loss Functions in Knowledge Graph Embedding Models

    Mohamed, Sameh K., Nov´ acek, V´ ıt, Vandenbussche, Pierre-Yves, and Mu˜ noz, Emir. “Loss Functions in Knowledge Graph Embedding Models”. In: DL4KG@ESWC. 2019. REFERENCES 219

    2019

  53. [62]

    A Re- view of Relational Machine Learning for Knowledge Graphs

    Nickel, Maximilian, Murphy, Kevin, Tresp, Volker, and Gabrilovich, Evgeniy. “A Re- view of Relational Machine Learning for Knowledge Graphs”. In: Proceedings of the IEEE 104.1 (2016), pp. 11–33. doi: 10.1109/JPROC.2015.2483592

    2016

  54. [63]

    BioPortal: ontologies and integrated data resources at the click of a mouse

    Noy, Natalya F, Shah, Nigam H, Whetzel, Patricia L, Dai, Benjamin, Dorf, Michael, Griffith, Nicholas, Jonquet, Clement, Rubin, Daniel L, Storey, Margaret-Anne, Chute, Christopher G, et al. “BioPortal: ontologies and integrated data resources at the click of a mouse”. In: Nucle...

    2009

  55. [64]

    Investigating Robustness and Interpretability of Link Prediction via Adversarial Modifications

    Pezeshkpour, Pouya, Tian, Yifan, and Singh, Sameer. “Investigating Robustness and Interpretability of Link Prediction via Adversarial Modifications”. In: Proceedings of the 2019 Conference of the North American Chapter of the Association for Computa- tional Linguistics: Human ...

    2019

  56. [65]

    The RDF2vec family of knowledge graph embed- ding methods

    Portisch, Jan and Paulheim, Heiko. “The RDF2vec family of knowledge graph embed- ding methods”. In: Semantic Web Preprint (2024), pp. 1–32

    2024

  57. [66]

    RNNLogic: Learning Logic Rules for Reasoning on Knowledge Graphs

    Qu, Meng, Chen, Junkun, Xhonneux, Louis-Pascal, Bengio, Yoshua, and Tang, Jian. “RNNLogic: Learning Logic Rules for Reasoning on Knowledge Graphs”. In: Interna- tional Conference on Learning Representations. 2020

    2020

  58. [67]

    Query2box: Reasoning Over Knowl- edge Graphs In Vector Space Using Box Embeddings

    Ren, Hongyu, Hu, Weihua, and Leskovec, Jure. “Query2box: Reasoning Over Knowl- edge Graphs In Vector Space Using Box Embeddings”. In: International Conference on Learning Representations (ICLR). 2020

    2020

  59. [68]

    RDF2Vec: RDF graph embeddings and their applications

    Ristoski, Petar, Rosati, Jessica, Di Noia, Tommaso, De Leone, Renato, and Paulheim, Heiko. “RDF2Vec: RDF graph embeddings and their applications”. In: Semantic Web 10.4 (2019), pp. 721–752

    2019

  60. [69]

    A neuro- symbolic system over knowledge graphs for link prediction

    Rivas, Ariam, Collarana, Diego, Torrente, Maria, and Vidal, Maria-Esther. “A neuro- symbolic system over knowledge graphs for link prediction”. In: Semantic Web 15.4 (2024), pp. 1307–1331

    2024

  61. [70]

    Knowledge Graph Embedding for Link Prediction: A Comparative Anal- ysis

    Rossi, Andrea, Barbosa, Denilson, Firmani, Donatella, Matinata, Antonio, and Meri- aldo, Paolo. “Knowledge Graph Embedding for Link Prediction: A Comparative Anal- ysis”. In: ACM Transactions on Knowledge Discovery from Data 15 (Jan. 2021), pp. 1–

    2021

  62. [71]

    Knowledge graph embeddings: Are relation- learning models learning relations?

    Rossi, Andrea and Matinata, Antonio. “Knowledge graph embeddings: Are relation- learning models learning relations?” In: EDBT/ICDT Workshops. Vol. 2578. 2020

    2020

  63. [72]

    The wisdom hierarchy: representations of the DIKW hierarchy

    Rowley, Jennifer. “The wisdom hierarchy: representations of the DIKW hierarchy”. In: Journal of information science 33.2 (2007), pp. 163–180. 220 REFERENCES

    2007

  64. [73]

    doi: 10.1145/3424672

    doi

  65. [74]

    Embedding knowledge graphs attentive to positional and centrality qualities

    Sadeghi, Afshin, Collarana, Diego, Graux, Damien, and Lehmann, Jens. “Embedding knowledge graphs attentive to positional and centrality qualities”. In: Joint European Conference on Machine Learning and Knowledge Discovery in Databases . Springer. 2021, pp. 548–564

    2021

  66. [75]

    CoDEx: A Comprehensive Knowledge Graph Com- pletion Benchmark

    Safavi, Tara and Koutra, Danai. “CoDEx: A Comprehensive Knowledge Graph Com- pletion Benchmark”. In: Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP). 2020

    2020

  67. [76]

    You CAN Teach an Old Dog New Tricks! On Training Knowledge Graph Embeddings

    Ruffinelli, Daniel, Broscheit, Samuel, and Gemulla, Rainer. “You CAN Teach an Old Dog New Tricks! On Training Knowledge Graph Embeddings”. In: ICLR. 2020

    2020

  68. [77]

    Veni, Vidi, Vici: Solving the Myriad of Challenges before Knowledge Graph Learning

    Sardina, Jeffrey, Costabello, Luca, and Gu´ eret, Christophe. “Veni, Vidi, Vici: Solving the Myriad of Challenges before Knowledge Graph Learning”. In: 2024 IEEE 18th International Conference on Semantic Computing (ICSC) . IEEE. 2024, pp. 197–203

    2024

  69. [78]

    NamE: Capturing Biological Context in KGEs via Contextual Named Graph Em- beddings

    Sardina, Jeffrey, Costabello, Luca, Gu´ eret, Christophe, and Mohammadi, Hossein. “NamE: Capturing Biological Context in KGEs via Contextual Named Graph Em- beddings”. In: 2024 46th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC) ...

    2024

  70. [79]

    Systems and methods for enhancing an artificial intelligence model via a multi-field embedding approach

    Sardina, Jeffrey, Bernardi, Alberto, and Gu´ eret, Christophe. Systems and methods for enhancing an artificial intelligence model via a multi-field embedding approach . US Patent App. 18/351,959. Jan. 2025

    2025

  71. [80]

    Extending TWIG: Zero- Shot Predictive Hyperparameter Selection for KGEs based on Graph Structure

    Sardina, Jeffrey, Kelleher, John D, and O’Sullivan, Declan. “Extending TWIG: Zero- Shot Predictive Hyperparameter Selection for KGEs based on Graph Structure”. In: Irish Conference on Artificial Intelligence and Cognitive Science . Springer. 2024

    2024

  72. [81]

    A Survey on Knowledge Graph Structure and Knowledge Graph Embeddings

    Sardina, Jeffrey, Kelleher, John D, and O’Sullivan, Declan. “A Survey on Knowledge Graph Structure and Knowledge Graph Embeddings”. In: 2025 IEEE 19th Interna- tional Conference on Semantic Computing (ICSC) . IEEE. 2025

    2025

  73. [82]

    TWIG-I: Embedding-Free Link Prediction and Cross-KG Transfer Learning Using a Small Neu- ral Architecture

    Sardina, Jeffrey, Debnath, Alok, Kelleher, John D, and O’Sullivan, Declan. “TWIG-I: Embedding-Free Link Prediction and Cross-KG Transfer Learning Using a Small Neu- ral Architecture”. In: Knowledge Graphs in the Age of Language Models and Neuro- Symbolic AI. IOS Press, 2024, p...

    2024

  74. [83]

    Structural Characteristics of Knowledge Gra-phs Determine the Quality of Knowledge Graph Embeddings Across Model and Hyperparameter Choices

    Sardina, Jeffrey and O’Sullivan, Declan. “Structural Characteristics of Knowledge Gra-phs Determine the Quality of Knowledge Graph Embeddings Across Model and Hyperparameter Choices”. In: SeWeBMeDA-2022@ESWC. 2022

    2022

  75. [84]

    Analysis of attention mechanisms in box-embedding systems

    Sardina, Jeffrey, Sardina, Callie, Kelleher, John D, and O’Sullivan, Declan. “Analysis of attention mechanisms in box-embedding systems”. In:Irish Conference on Artificial Intelligence and Cognitive Science . Springer. 2022, pp. 68–80

    2022

  76. [85]

    TWIG: Towards pre-hoc Hyperparameter Optimisation and Cross-Graph Generalisation via Simulated KGE Models

    Sardina, Jeffrey, Kelleher, John D, and O’Sullivan, Declan. “TWIG: Towards pre-hoc Hyperparameter Optimisation and Cross-Graph Generalisation via Simulated KGE Models”. In: 2024 IEEE 18th International Conference on Semantic Computing (IC- SC). IEEE. 2024, pp. 122–129. REFERENCES 221

    2024

  77. [86]

    Combining rules and embeddings via neuro-symbolic ai for knowledge base completion

    Sen, Prithviraj, Carvalho, Breno WSR, Abdelaziz, Ibrahim, Kapanipathi, Pavan, Luus, Francois, Roukos, Salim, and Gray, Alexander. “Combining rules and embeddings via neuro-symbolic ai for knowledge base completion”. In: arXiv preprint arXiv:2109.- 09566 (2021)

    2021

  78. [87]

    Dif- ferentiable neuro-symbolic reasoning on large-scale knowledge graphs

    Shengyuan, Chen, Cai, Yunfeng, Fang, Huang, Huang, Xiao, and Sun, Mingming. “Dif- ferentiable neuro-symbolic reasoning on large-scale knowledge graphs”. In: Advances in Neural Information Processing Systems 36 (2023), pp. 28139–28154

    2023

  79. [88]

    Modeling relational data with graph convolutional net- works

    Schlichtkrull, Michael, Kipf, Thomas N, Bloem, Peter, Van Den Berg, Rianne, Titov, Ivan, and Welling, Max. “Modeling relational data with graph convolutional net- works”. In: The semantic web: 15th international conference, ESWC 2018, Heraklion, Crete, Greece, June 3–7, 2018, ...

    2018

  80. [89]

    Applying network link prediction in drug discovery: an overview of the literature

    Son, Jeongtae and Kim, Dongsup. “Applying network link prediction in drug discovery: an overview of the literature”. In: Expert Opinion on Drug Discovery 19.1 (2024), pp. 43–56

    2024

  81. [90]

    Conceptnet 5.5: An open mul- tilingual graph of general knowledge

    Speer, Robyn, Chin, Joshua, and Havasi, Catherine. “Conceptnet 5.5: An open mul- tilingual graph of general knowledge”. In: Proceedings of the AAAI conference on artificial intelligence. Vol. 31. 1. 2017

    2017

  82. [91]

    Open-world knowledge graph completion

    Shi, Baoxu and Weninger, Tim. “Open-world knowledge graph completion”. In: Pro- ceedings of the AAAI conference on artificial intelligence . Vol. 32. 1. 2018

    2018

  83. [92]

    The Fellowship of the Ring

    Tolkien, John Ronald Reuel. The Fellowship of the Ring . HarperCollins Publishers, 2008

    2008

  84. [93]

    Observed versus latent features for knowl- edge base and text inference

    Toutanova, Kristina and Chen, Danqi. “Observed versus latent features for knowl- edge base and text inference”. In: Proceedings of the 3rd Workshop on Continuous Vector Space Models and their Compositionality. Ed. by Alexandre Allauzen, Edward Grefenstette, Karl Moritz Hermann...

    2015 doi

  85. [94]

    A benchmarking study of embedding-based entity align- ment for knowledge graphs

    Sun, Zequn, Zhang, Qingheng, Hu, Wei, Wang, Chengming, Chen, Muhao, Akrami, Farahnaz, and Li, Chengkai. “A benchmarking study of embedding-based entity align- ment for knowledge graphs”. In: Proceedings of the VLDB Endowment 13.12 (2020)

    2020

  86. [95]

    Compo- sition-based Multi-Relational Graph Convolutional Networks

    Vashishth, Shikhar, Sanyal, Soumya, Nitin, Vikram, and Talukdar, Partha. “Compo- sition-based Multi-Relational Graph Convolutional Networks”. In: International Con- ference on Learning Representations. 2019

    2019

  87. [96]

    A survey on knowledge graph em- beddings for link prediction

    Wang, Meihong, Qiu, Linling, and Wang, Xiaoli. “A survey on knowledge graph em- beddings for link prediction”. In: Symmetry 13.3 (2021), p. 485

    2021

  88. [97]

    Complex embeddings for simple link prediction

    Trouillon, Th´ eo, Welbl, Johannes, Riedel, Sebastian, Gaussier, ´Eric, and Bouchard, Guillaume. “Complex embeddings for simple link prediction”. In: International con- ference on machine learning . PMLR. 2016, pp. 2071–2080

    2016

  89. [98]

    Knowledge graph embedding by translating on hyperplanes

    Wang, Zhen, Zhang, Jianwen, Feng, Jianlin, and Chen, Zheng. “Knowledge graph embedding by translating on hyperplanes”. In: Proceedings of the AAAI conference on artificial intelligence . Vol. 28. 1. 2014

    2014

  90. [99]

    DrugBank 5.0: a major update to the DrugBank database for 2018

    Wishart, David S, Feunang, Yannick D, Guo, An C, Lo, Elvis J, Marcu, Ana, Grant, Jason R, Sajed, Tanvir, Johnson, Daniel, Li, Carin, Sayeeda, Zinat, et al. “DrugBank 5.0: a major update to the DrugBank database for 2018”. In: Nucleic acids research 46.D1 (2018), pp. D1074–D1082

    2018

  91. [100]

    Knowledge Graph Embed- ding: A Survey of Approaches and Applications

    Wang, Quan, Mao, Zhendong, Wang, Bin, and Guo, Li. “Knowledge Graph Embed- ding: A Survey of Approaches and Applications”. In: IEEE Transactions on Knowl- edge and Data Engineering 29.12 (2017), pp. 2724–2743. doi: 10.1109/TKDE.2017. 2754499

    2017 doi

  92. [101]

    OpenGraph: Towards Open Graph Foundation Models

    Xia, Lianghao, Kao, Ben, and Huang, Chao. “OpenGraph: Towards Open Graph Foundation Models”. In: Findings of the Association for Computational Linguistics: EMNLP 2024. 2024, pp. 2365–2379

    2024

  93. [102]

    Em- bedding Entities and Relations for Learning and Inference in Knowledge Bases

    Yang, Bishan, Yih, Scott Wen-tau, He, Xiaodong, Gao, Jianfeng, and Deng, Li. “Em- bedding Entities and Relations for Learning and Inference in Knowledge Bases”. In: Proceedings of the International Conference on Learning Representations (ICLR)

  94. [103]

    An on- tology model for climatic data analysis

    Wu, Jiantao, Orlandi, Fabrizio, O’Sullivan, Declan, and Dev, Soumyabrata. “An on- tology model for climatic data analysis”. In: 2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS . IEEE. 2021, pp. 5739–5742

    2021

  95. [104]

    Data poisoning attack against knowledge graph embedding

    Zhang, Hengtong, Zheng, Tianhang, Gao, Jing, Miao, Chenglin, Su, Lu, Li, Yaliang, and Ren, Kui. “Data poisoning attack against knowledge graph embedding”. In: Pro- ceedings of the 28th International Joint Conference on Artificial Intelligence . 2019, pp. 4853–4859

    2019

  96. [105]

    Drug repurposing for COVID-19 via knowledge graph comple- tion

    Zhang, Rui, Hristovski, Dimitar, Schutte, Dalton, Kastrin, Andrej, Fiszman, Marcelo, and Kilicoglu, Halil. “Drug repurposing for COVID-19 via knowledge graph comple- tion”. In: Journal of biomedical informatics 115 (2021), p. 103696

    2021

  97. [106]

    ADADELTA: an adaptive learning rate method

    Zeiler, Matthew D. “ADADELTA: an adaptive learning rate method”. In: arXiv pre- print arXiv:1212.5701 (2012). REFERENCES 223

    2012 arXiv

  98. [107]

    Iteratively learning embeddings and rules for knowledge graph reasoning

    Zhang, Wen, Paudel, Bibek, Wang, Liang, Chen, Jiaoyan, Zhu, Hai, Zhang, Wei, Bern- stein, Abraham, and Chen, Huajun. “Iteratively learning embeddings and rules for knowledge graph reasoning”. In: The world wide web conference. 2019, pp. 2366–2377

    2019

  99. [108]

    PharmKG: a dedicated knowledge graph benchmark for bomedical data mining

    Zheng, Shuangjia, Rao, Jiahua, Song, Ying, Zhang, Jixian, Xiao, Xianglu, Fang, Evan- dro Fei, Yang, Yuedong, and Niu, Zhangming. “PharmKG: a dedicated knowledge graph benchmark for bomedical data mining”. In: Briefings in bioinformatics 22.4 (2021), bbaa344

    2021

  100. [109]

    Knowledge graph reasoning with logics and embeddings: Survey and perspective

    Zhang, Wen, Chen, Jiaoyan, Li, Juan, Xu, Zezhong, Pan, Jeff Z, and Chen, Huajun. “Knowledge graph reasoning with logics and embeddings: Survey and perspective”. In: 2024 IEEE International Conference on Knowledge Graph (ICKG) . IEEE. 2024, pp. 492–499

    2024

  101. [110]

    The probability of edge existence due to node degree: a baseline for network-based predictions

    Zietz, Michael, Himmelstein, Daniel S, Kloster, Kyle, Williams, Christopher, Nagle, Michael W, and Greene, Casey S. “The probability of edge existence due to node degree: a baseline for network-based predictions”. In: GigaScience 13 (2024), giae001

    2024

  102. [112]

    Graph neural networks: A review of methods and applications

    Zhou, Jie, Cui, Ganqu, Hu, Shengding, Zhang, Zhengyan, Yang, Cheng, Liu, Zhiyuan, Wang, Lifeng, Li, Changcheng, and Sun, Maosong. “Graph neural networks: A review of methods and applications”. In: AI open 1 (2020), pp. 57–81

    2020

  103. [300]

    url: https://doi.org/10.1145/3447772

    doi: 10.1145/3447772. url: https://doi.org/10.1145/3447772

    doi

  104. [2006]

    Springer

    Proceedings 17. Springer. 2006, pp. 98–111

    2006

Pith tools

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