Pith. sign in

REVIEW 4 major objections 4 minor 40 references

Domain and Range Aware Synthetic Negatives Generation for Knowledge Graph Embedding Models

T0 review · 4 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

Pith's one-line read Domain-aware negative sampling lifts knowledge-graph link prediction by up to 150%.

desk verdict A simple and plausible negative-sampling interpolation, but the paper's own numbers are internally contradictory and the headline Hetionet gain rests on a split the authors concede is problematic. read the letter →

arxiv 2411.14858 v1 pith:Y3ZLHE73 submitted 2024-11-22 cs.AI

classification cs.AI
keywords knowledgegraphembeddingnegativesamplinglinkpredictiontypeconstraintsdomainandrangeontologyHetionetFB15k-237
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 argues that synthetic negative triples for training knowledge graph embedding models should be drawn neither purely at random nor purely from a relation's type constraints, but from a mixture of the two. The authors extend an existing type-constrained strategy that corrupts triples by replacing the head or tail with entities from the relation's domain or range, and blend it with uniform random sampling through a hyperparameter $\nu$ that sets the proportion of each. On FB15k-237 the mixture outperforms both baselines across all four models tested, with up to +10% mean reciprocal rank (MRR) over random and +33% over type-constrained negative sampling; on the ontology-backed Hetionet dataset the relative gain over random reaches more than +150% MRR. The only exceptions are TransE and DistMult on WN18RR, where the mixture does not beat random sampling. If the result holds, a nearly free change to how negatives are generated can materially improve link prediction, and it can do so with fewer negatives per triple.

What carries the argument

The load-bearing object is the negative set constructed for each training triple $t=(s,p,o)$. Type-constrained negatives are drawn from $T^-_{t.c.} = \{(s',p,o) \mid s' \in \mathrm{domain}_p\} \cup \{(s,p,o') \mid o' \in \mathrm{range}_p\}$; random negatives come from replacing either entity with any entity in the graph. The method partitions the $\eta$ negatives per triple into $\lfloor\eta\nu\rfloor$ random and $\lceil\eta(1-\nu)\rceil$ type-constrained samples, so $\nu$ controls the trade-off between semantic relevance and diversity. This addresses the two failure modes of the pure strategies: pure random sampling is dominated by over-represented entity types, while pure type-constrained sampling repeats the same entities and produces false negatives when a class has very few instances.

What would settle it

Rerun the Hetionet experiments with a non-random held-out split (e.g., withholding triples by relation type or by connected component) and with several random seeds, then check whether the mixture still beats random sampling by a large margin; if the >150% MRR advantage does not persist, the paper's strongest empirical claim is not supported.

Watch

Extended reading notes

Core claim

The central claim is that domain- and range-aware negative sampling, when combined with random uniform sampling, improves link prediction for knowledge graph embedding models relative to either strategy alone. The paper defines, for each relation $p$, the domain as all entities that appear as a subject in some $(s,p,o)$ triple and the range as all entities that appear as an object; type-constrained negatives corrupt a triple only within those sets. The proposed method samples a fraction $\nu$ of the $\eta$ negatives from this constrained set and the remaining fraction from the full entity set. In experiments with TransE, DistMult, ComplEx-N3, and RotatE, this mixture beats the random and type-constrained baselines in every dataset except WN18RR with TransE and DistMult, where it falls slightly short of the random baseline. On Hetionet, where classes come from a real ontology and some types such as Gene dominate, the gains are largest: well over +150% MRR relative to random sampling.

Load-bearing premise

The crux is the assumption that the relative ordering of negative-sampling strategies measured on a random 10,000-triple test split of Hetionet carries over to the real setting in which missing edges are not missing at random; on top of that, all results come from a single run, so the size of the reported gains has no measured variance.

Editorial extensions

If this is right

  • On FB15k-237, the mixture gives a 10% MRR improvement over random sampling for ComplEx-N3 and a 33% improvement over type-constrained negative sampling, without changing the model architecture.
  • On ontology-backed graphs with skewed class sizes, such as Hetionet, the strategy can multiply MRR by roughly 2.5 relative to random sampling while adding only about 5 ms per training step.
  • The benefit is most pronounced when the number of negatives per triple is small ($\eta=10$); with larger $\eta$, random sampling alone supplies enough informative negatives, so the strategy can train with fewer samples.
  • The optimal $\nu$ is easy to tune: results are consistent across neighbouring values and across all four scoring functions on a given dataset.

Reading between the lines

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

  • The large Hetionet improvement may not survive a more realistic evaluation: the paper uses a random split of triples, whereas in practice missing edges are not missing at random; a systematic split by relation or by entity subset would test this directly.
  • The method's value should depend on the cardinality distribution of domain and range classes; a diagnostic that varies class sizes while holding the graph fixed could predict when the mixture helps and by how much.
  • Because the strategy is scoring-function agnostic in the paper's experiments, it is plausible to combine it with harder-negative or self-adversarial schemes, but that extension is untested and goes beyond what the paper shows.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper proposes a negative-sampling strategy for knowledge graph embedding (KGE) models that mixes type-constrained negatives (drawn from the domain or range of the relation) with uniform-random negatives, with the mixture ratio controlled by a hyperparameter ν. The method is evaluated on FB15k-237, WN18RR, and Hetionet with TransE, DistMult, ComplEx-N3, and RotatE, comparing against uniform-random and pure type-constrained baselines. The authors report consistent improvements on FB15k-237 across all four models, mixed results on WN18RR, and very large relative gains on Hetionet, and claim that the strategy is simple, scalable, and leads to state-of-the-art-competitive performance.

Significance. If the empirical claims hold, the paper provides a deceptively simple modification to negative sampling that could improve KGE link prediction on ontology-rich graphs, with little computational overhead. The strengths include a systematic grid search over hyperparameters, evaluation across four model families and three datasets, and a clear pseudo-code description of the method. However, the headline gains rest on a single evaluation protocol for Hetionet that the authors themselves flag as problematic, and the abstract's '+10% MRR' is a best-case number from one model on one dataset rather than a representative result. The work is potentially useful to practitioners, but the current evidence does not fully support the broad claims.

major comments (4)
  1. [Section 4.1, footnote 1; Table 1 (Hetionet rows)] The central claim of over +150% MRR on Hetionet rests entirely on a random 10,000-triple test split that the authors concede 'poses multiple challenges.' Under a random split, test triples are missing uniformly at random from the same distribution used to train, which can systematically inflate the apparent value of type-constrained negatives because the domain/range constraints used at test time align tightly with the training-time corruption distribution. The authors state that their purpose is a relative comparison of negative-sampling strategies, but this is precisely the assumption that needs to be verified: the relative ranking under random missingness does not necessarily transfer to realistic settings where missing edges are not missing at random. Since no variance estimates or multiple-seed runs are reported, it is also unclear whether the large margins on Hetionet (e.g., RotatE 0.26 vs 0.10 MRR) are stable. I ask the authors to evaluate at least one non-random split, such as the disease-centric split of [31], and to report standard deviations over multiple seeds before the Hetionet claim can be accepted.
  2. [Appendix A.5 vs Table 1 (TransE on WN18RR)] There is a direct internal inconsistency that undermines the credibility of the WN18RR results. Appendix A.5 states that 'the MRR of TransE on WN18RR ... was in the range of [0.03, 0.04] for all νs, including the baseline,' but Table 1 lists TransE with random sampling at 0.25 MRR, type-constrained at 0.19, and the proposed method at 0.24. These two numbers cannot both be correct. Since the paper's own narrative relies on TransE (and DistMult) being the exceptions on WN18RR, this discrepancy must be resolved—either the table or the appendix text is wrong, and the correct values must be reported.
  3. [Section 3.2, Algorithm 1] The pseudo-code in Algorithm 1 contains a typo that makes the algorithm ill-defined as written: the line 'St.c. ← sample(T − t.c., ηrand)' uses ηrand instead of ηt.c. for the number of type-constrained negatives. With this line, when ν=0 (pure random baseline), ηrand=0 and no negatives would be generated at all, contradicting the experiments where ν=0 is the random baseline. The implementation presumably uses ηt.c., but the paper must correct the pseudo-code, as it is the formal specification of the proposed method.
  4. [Abstract and Section 4.2] The abstract's claim of 'substantial improvement (+10% MRR) for standard benchmark datasets' is cherry-picked: the +10% figure is the largest gain on FB15k-237 (ComplEx-N3), while WN18RR shows no improvement for DistMult and ComplEx-N3 and a regression for TransE under the proposed method. The abstract should either present the full range of results or state the average improvement, rather than highlighting only the best case. This overstates the strength and consistency of the empirical evidence.
minor comments (4)
  1. [Section 4.2] The text states that 'the optimal value of η for TransE, DistMult and ComplEx-N3 on WN18RR is 30,' but Table 3 shows that ComplEx-N3 achieves best results with η=20 on WN18RR for all three negative-sampling strategies. This should be corrected or clarified.
  2. [Table 1 caption] The caption says 'Best results for each datasets are reported in bold, while across models are underlined,' but the table as rendered does not contain any bold or underlined entries. The formatting should be made consistent or the caption removed.
  3. [Appendix A.5, Figure 1] Figure 1 is referenced and captioned, but the actual plot is not visible in the provided text. Please ensure the figure is included in the final version.
  4. [Section 3.1-3.2] The definition of domainp and rangep in Section 3.1 is given in terms of the graph G, but Section 3.2 says the type constraints are 'extracted from the ontology (if given) or from the facts in the training data.' For Hetionet, where an ontology is available, it is unclear whether the domain/range sets are taken from the ontology or inferred from the training facts. This distinction matters for the interpretation of the Hetionet results and should be made explicit.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the proposed negative-sampling strategy is validated on held-out test sets against external baselines; ν is tuned on validation, and the only self-citation (AmpliGraph library) is an implementation tool, not load-bearing.

full rationale

The paper's derivation chain is empirical rather than analytic: the method takes the type-constrained negative set T^{-}_{t.c.} defined in Section 3.2 and mixes it with random negatives via hyperparameter ν, then measures MRR/Hits@N on held-out test splits for FB15k-237, WN18RR, and Hetionet. No claimed prediction is constructed from the evaluation data: ν is selected by grid search on validation (Section 4.1), the baselines are external strategies ([3] random, [7] type-constrained), and the domain/range classes are defined from training facts or from the Hetionet ontology rather than from test labels. The one self-citation, AmpliGraph [38], is the software library used to run experiments; it is a tool citation and does not supply the theoretical premise or the comparisons, so it is not load-bearing. The flagged Hetionet random-split limitation (footnote 1) is an acknowledged evaluation-validity concern (missingness is not at random), and the Appendix A.5 vs Table 1 TransE WN18RR MRR discrepancy is an internal reporting inconsistency; both are correctness/robustness issues, not circular reductions. Because the benchmark results could, in principle, fail to replicate without invalidating the derivation, the central claim is not equivalent to its inputs by construction.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The new contribution rests on one invented hyperparameter (nu), standard training hyperparameters, and two domain assumptions that carry most of the risk, namely type-set informativeness and the Hetionet split. The contribution introduces no new entities, forces, or mathematical objects.

free parameters (3)
  • nu (mixture proportion between random and type-constrained negatives) = FB15k-237: 0.1-0.2; WN18RR: 0.2-0.8; Hetionet: 0.3-0.8 (Appendix A.4)
    Introduced by this paper as the core new hyperparameter. Selected per dataset and model by grid search; the reported improvements are conditional on this tuning.
  • eta (number of negatives per triple) = 10, 20, or 30 depending on model and dataset (Appendix A.4)
    Standard negative-sampling hyperparameter; tuned in the grid. The claim that the method allows fewer negatives is model-dependent.
  • Embedding dimension k = 200 or 350 (Appendix A.4)
    Standard KGE hyperparameter; limited to smaller sizes to keep the comparison fair against expensive SOTA models.
assumptions (4)
  • domain assumption Domain and range sets of each relation, derived from training facts or ontology, provide a valid pool for generating semantically meaningful negatives.
    This is the premise of type-constrained sampling (Krompaß et al. [7]); the paper relies on it and extends it. If these sets are noisy, the type-constrained component is misinformed.
  • domain assumption A random 10,000-triple test split of Hetionet is a fair basis for comparing negative sampling strategies.
    The authors note in Section 4.1 (footnote 1) that properly splitting Hetionet is challenging; the +150% gain claim is measured on this split.
  • domain assumption A single training run per configuration is representative enough to compare methods.
    No multiple seeds are reported; observed 0.01-0.02 MRR differences are treated as meaningful.
  • standard math Filtered setting evaluation with all-entity corruption is the standard link prediction protocol.
    Standard in the KGE literature and used without modification.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Domain and Range Aware Synthetic Negatives Generation for Knowledge Graph Embedding Models." pith.science (2026). https://pith.science/paper/Y3ZLHE73

@misc{pith2026241114858,
  author       = {Pith},
  title        = {Pith review of: Domain and Range Aware Synthetic Negatives Generation for Knowledge Graph Embedding Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Y3ZLHE73}},
  note         = {Machine review of arXiv:2411.14858}
}
read the original abstract

Knowledge Graph Embedding models, representing entities and edges in a low-dimensional space, have been extremely successful at solving tasks related to completing and exploring Knowledge Graphs (KGs). One of the key aspects of training most of these models is teaching to discriminate between true statements positives and false ones (negatives). However, the way in which negatives can be defined is not trivial, as facts missing from the KG are not necessarily false and a set of ground truth negatives is hardly ever given. This makes synthetic negative generation a necessity. Different generation strategies can heavily affect the quality of the embeddings, making it a primary aspect to consider. We revamp a strategy that generates corruptions during training respecting the domain and range of relations, we extend its capabilities and we show our methods bring substantial improvement (+10% MRR) for standard benchmark datasets and over +150% MRR for a larger ontology-backed dataset.

Figures

Figures reproduced from arXiv: 2411.14858 by the authors.

Figure 1
Figure 1. Performance of the different models varying the proportion of ontology-based negatives ν. types. In FB15k-237 and WN18RR the improvement is less marked as there is no class significantly over-represented. On the other side of the spectrum of values for ν, we can observe that some models see a significant decrease of the performance as we approach the Type-Constrained baseline (ν = 1). This is definitely dataset-depe… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

40 extracted references · 21 canonical work pages

  1. [31]

    Understanding the Performance of Knowledge Graph Embeddings in Drug Discovery

    Stephen Bonner, Ian P. Barrett, Cheng Ye, Rowan Swiers, Ola Engkvist, and William L. Hamilton. Understanding the performance of knowledge graph embeddings in drug discovery. CoRR, abs/2105.10488, 2021. URL https://arxiv.org/abs/2105.10488. 3

  2. [1]

    Rashid, Anisa Rula, Lukas Schmelzeisen, Juan Sequeda, Steffen Staab, and Antoine Zimmermann

    Aidan Hogan, Eva Blomqvist, Michael Cochez, Claudia D’amato, Gerard De Melo, Claudio Gutierrez, Sabrina Kirrane, José Emilio Labra Gayo, Roberto Navigli, Sebastian Neumaier, Axel-Cyrille Ngonga Ngomo, Axel Polleres, Sabbir M. Rashid, Anisa Rula, Lukas Schmelzeisen, Juan Sequeda, Steffen Staab, and Antoine Zimmermann. Knowledge graphs. 54(4), jul 2021. ISS...

  3. [2]

    Knowledge vault: a web-scale approach to probabilis- tic knowledge fusion

    Xin Dong, Evgeniy Gabrilovich, Geremy Heitz, Wilko Horn, Ni Lao, Kevin Murphy, Thomas Strohmann, Shaohua Sun, and Wei Zhang. Knowledge vault: a web-scale approach to probabilis- tic knowledge fusion. In Proceedings of the 20th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD ’14, page 601–610, New York, NY , USA, 2014. Assoc...

  4. [3]

    Translating embeddings for modeling multi-relational data

    Antoine Bordes, Nicolas Usunier, Alberto Garcia-Duran, Jason Weston, and Oksana Yakhnenko. Translating embeddings for modeling multi-relational data. In C.J. Burges, L. Bottou, M. Welling, Z. Ghahramani, and K.Q. Weinberger, editors, Ad- vances in Neural Information Processing Systems , volume 26. Curran Associates, Inc.,

  5. [4]

    Factorizing Y AGO: scalable machine learning for linked data

    Maximilian Nickel, V olker Tresp, and Hans-Peter Kriegel. Factorizing Y AGO: scalable machine learning for linked data. In Alain Mille, Fabien Gandon, Jacques Misselis, Michael Rabinovich, and Steffen Staab, editors, Proceedings of the 21st World Wide Web Conference 2012, WWW 2012, Lyon, France, April 16-20, 2012, pages 271–280. ACM, 2012. doi: 10.1145/21...

  6. [5]

    Comprehensive analysis of negative sampling in knowledge graph representation learning

    Hidetaka Kamigaito and Katsuhiko Hayashi. Comprehensive analysis of negative sampling in knowledge graph representation learning. In Kamalika Chaudhuri, Stefanie Jegelka, Le Song, Csaba Szepesvári, Gang Niu, and Sivan Sabato, editors, International Conference on Ma- chine Learning, ICML 2022, 17-23 July 2022, Baltimore, Maryland, USA , volume 162 of Proce...

  7. [6]

    Negative sampling in knowledge graph representation learning: A review

    Tiroshan Madushanka and Ryutaro Ichise. Negative sampling in knowledge graph representation learning: A review. CoRR, abs/2402.19195, 2024. doi: 10.48550/ARXIV .2402.19195. URL https://doi.org/10.48550/arXiv.2402.19195. 1, 2

  8. [7]

    Type-constrained representation learning in knowledge graphs

    Denis Krompaß, Stephan Baier, and V olker Tresp. Type-constrained representation learning in knowledge graphs. In Marcelo Arenas, Óscar Corcho, Elena Simperl, Markus Strohmaier, Mathieu d’Aquin, Kavitha Srinivas, Paul Groth, Michel Dumontier, Jeff Heflin, Krishnaprasad Thirunarayan, and Steffen Staab, editors, The Semantic Web - ISWC 2015 - 14th Internati...

Show all 40 references
  1. [8]

    You CAN teach an old dog new tricks! on training knowledge graph embeddings

    Daniel Ruffinelli, Samuel Broscheit, and Rainer Gemulla. You CAN teach an old dog new tricks! on training knowledge graph embeddings. In 8th International Conference on Learning Representations, ICLR 2020, Addis Ababa, Ethiopia, April 26-30, 2020. OpenReview.net, 2020. URL htt...

  2. [9]

    Convolutional 2d knowledge graph embeddings

    Tim Dettmers, Pasquale Minervini, Pontus Stenetorp, and Sebastian Riedel. Convolutional 2d knowledge graph embeddings. In Sheila A. McIlraith and Kilian Q. Weinberger, editors, Proceedings of the Thirty-Second AAAI Conference on Artificial Intelligence, (AAAI-18), the 30th inn...

  3. [10]

    Knowledge graph embedding: A survey from the perspective of representation spaces

    Jiahang Cao, Jinyuan Fang, Zaiqiao Meng, and Shangsong Liang. Knowledge graph embedding: A survey from the perspective of representation spaces. ACM Comput. Surv., 56(6), mar 2024. ISSN 0360-0300. doi: 10.1145/3643806. URL https://doi.org/10.1145/3643806. 2

  4. [11]

    Investigations on knowledge base embedding for relation prediction and extraction, 2018

    Peng Xu and Denilson Barbosa. Investigations on knowledge base embedding for relation prediction and extraction, 2018. 2 5 Domain and Range Aware Negatives for KGE

  5. [12]

    Pytorch-biggraph: A large scale graph embedding system

    Adam Lerer, Ledell Wu, Jiajun Shen, Timothée Lacroix, Luca Wehrstedt, Abhijit Bose, and Alex Peysakhovich. Pytorch-biggraph: A large scale graph embedding system. In Ameet Talwalkar, Virginia Smith, and Matei Zaharia, editors,Proceedings of the Second Conference on Machine Lea...

  6. [13]

    Knowledge graph embedding by translating on hyperplanes

    Zhen Wang, Jianwen Zhang, Jianlin Feng, and Zheng Chen. Knowledge graph embedding by translating on hyperplanes. In Carla E. Brodley and Peter Stone, editors, Proceedings of the Twenty-Eighth AAAI Conference on Artificial Intelligence, July 27 -31, 2014, Québec City, Québec, C...

  7. [14]

    Qizhe Xie, Xuezhe Ma, Zihang Dai, and Eduard H. Hovy. An interpretable knowledge trans- fer model for knowledge base completion. In Regina Barzilay and Min-Yen Kan, editors, Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, ACL 2017, Vanc...

  8. [15]

    A novel negative sample generating method for knowledge graph embedding

    Yi Zhang, Wanhua Cao, and Juntao Liu. A novel negative sample generating method for knowledge graph embedding. In Yunhao Liu, Guoliang Xing, Yuan He, and Gian Pietro Picco, editors, Proceedings of the 2019 International Conference on Embedded Wireless Systems and Networks, EWS...

  9. [16]

    A novel negative sampling based on frequency of relational association entities for knowledge graph embedding

    Wanhua Cao, Yi Zhang, Juntao Liu, and Ziyun Rao. A novel negative sampling based on frequency of relational association entities for knowledge graph embedding. J. Web Eng., 20(6), 2021. doi: 10.13052/JWE1540-9589.2068. URL https://doi.org/10.13052/ jwe1540-9589.2068. 2

  10. [17]

    Enhancing knowledge graph embedding with probabilistic negative sampling

    Vibhor Kanojia, Hideyuki Maeda, Riku Togashi, and Sumio Fujita. Enhancing knowledge graph embedding with probabilistic negative sampling. In Rick Barrett, Rick Cummings, Eugene Agichtein, and Evgeniy Gabrilovich, editors, Proceedings of the 26th International Conference on Wor...

  11. [18]

    Analysis of the impact of negative sampling on link prediction in knowledge graphs

    Bhushan Kotnis and Vivi Nastase. Analysis of the impact of negative sampling on link prediction in knowledge graphs. CoRR, abs/1708.06816, 2017. URL http://arxiv.org/abs/1708. 06816. 2

  12. [19]

    Rashad Al Hasan Rony, Semab Ali, Jens Lehmann, and Sahar Vahdati

    Mirza Mohtashim Alam, Md. Rashad Al Hasan Rony, Semab Ali, Jens Lehmann, and Sahar Vahdati. Language model-driven negative sampling. CoRR, abs/2203.04703, 2022. doi: 10.48550/ARXIV .2203.04703. URLhttps://doi.org/10.48550/arXiv.2203.04703. 2

  13. [20]

    Yuncheng Wang, Xiou Ge, Bin Wang, and C.-C. Jay Kuo. Kgboost: A classification-based knowledge base completion method with negative sampling. Pattern Recognit. Lett., 157: 104–111, 2022. doi: 10.1016/J.PATREC.2022.04.001. URL https://doi.org/10.1016/j. patrec.2022.04.001. 2

  14. [21]

    Incorporating domain and range of relations for knowledge graph completion

    Juan Li, Wen Zhang, and Huajun Chen. Incorporating domain and range of relations for knowledge graph completion. In Xiaoyan Zhu, Bing Qin, Xiaodan Zhu, Ming Liu, and Longhua Qian, editors, Knowledge Graph and Semantic Computing: Knowledge Computing and Language Understanding -...

  15. [22]

    Conditional constraints for knowledge graph embeddings

    Michael Weyns, Pieter Bonte, Bram Steenwinckel, Filip De Turck, and Femke Ongenae. Conditional constraints for knowledge graph embeddings. In Mehwish Alam, Davide Buscaldi, Michael Cochez, Francesco Osborne, Diego Reforgiato Recupero, and Harald Sack, editors, Proceedings of t...

  16. [23]

    Rotate: Knowledge graph em- bedding by relational rotation in complex space

    Zhiqing Sun, Zhi-Hong Deng, Jian-Yun Nie, and Jian Tang. Rotate: Knowledge graph em- bedding by relational rotation in complex space. In 7th International Conference on Learning Representations, ICLR 2019, New Orleans, LA, USA, May 6-9, 2019. OpenReview.net, 2019. URL https://...

  17. [24]

    Confidence-aware negative sampling method for noisy knowledge graph embedding

    Yingchun Shan, Chenyang Bu, Xiaojian Liu, Shengwei Ji, and Lei Li. Confidence-aware negative sampling method for noisy knowledge graph embedding. In Xindong Wu, Yew-Soon Ong, Charu C. Aggarwal, and Huanhuan Chen, editors, 2018 IEEE International Conference on Big Knowledge, IC...

  18. [25]

    Nscaching: Simple and efficient negative sampling for knowledge graph embedding

    Yongqi Zhang, Quanming Yao, Yingxia Shao, and Lei Chen. Nscaching: Simple and efficient negative sampling for knowledge graph embedding. In 35th IEEE International Conference on Data Engineering, ICDE 2019, Macao, China, April 8-11, 2019, pages 614–625. IEEE, 2019. doi: 10.110...

  19. [26]

    Adversarial knowledge representation learning without external model

    Jingpei Lei, Dantong Ouyang, and Ying Liu. Adversarial knowledge representation learning without external model. IEEE Access, 7:3512–3524, 2019. doi: 10.1109/ACCESS.2018. 2889481. URL https://doi.org/10.1109/ACCESS.2018.2889481. 2

  20. [27]

    Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron C

    Ian J. Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron C. Courville, and Yoshua Bengio. Generative adversarial nets. In Zoubin Ghahramani, Max Welling, Corinna Cortes, Neil D. Lawrence, and Kilian Q. Weinberger, editors, Advances ...

  21. [28]

    Reconciling competing sampling strategies of network embedding

    Yuchen Yan, Baoyu Jing, Lihui Liu, Ruijie Wang, Jinning Li, Tarek Abdelzaher, and Hang- hang Tong. Reconciling competing sampling strategies of network embedding. In A. Oh, T. Naumann, A. Globerson, K. Saenko, M. Hardt, and S. Levine, editors, Advances in Neural Information Pr...

  22. [29]

    Observed versus latent features for knowledge base and text inference

    Kristina Toutanova and Danqi Chen. Observed versus latent features for knowledge base and text inference. In Alexandre Allauzen, Edward Grefenstette, Karl Moritz Hermann, Hugo Larochelle, and Scott Wen-tau Yih, editors,Proceedings of the 3rd Workshop on Continuous Vector Space...

  23. [30]

    Systematic integration of biomedical knowledge prioritizes drugs for repurposing

    Daniel Scott Himmelstein, Antoine Lizee, Christine Hessler, Leo Brueggeman, Sabrina L Chen, Dexter Hadley, Ari Green, Pouya Khankhanian, and Sergio E Baranzini. Systematic integration of biomedical knowledge prioritizes drugs for repurposing. eLife, 6:e26726, sep 2017. ISSN 20...

  24. [32]

    Kingma and Jimmy Ba

    Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization. In Yoshua Bengio and Yann LeCun, editors, 3rd International Conference on Learning Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings, 2015. URL http://arxiv.o...

  25. [33]

    Embedding entities and relations for learning and inference in knowledge bases

    Bishan Yang, Wen-tau Yih, Xiaodong He, Jianfeng Gao, and Li Deng. Embedding entities and relations for learning and inference in knowledge bases. In Yoshua Bengio and Yann LeCun, editors, 3rd International Conference on Learning Representations, ICLR 2015, San Diego, CA, USA, ...

  26. [34]

    Complex embeddings for simple link prediction

    Théo Trouillon, Johannes Welbl, Sebastian Riedel, Éric Gaussier, and Guillaume Bouchard. Complex embeddings for simple link prediction. In ICML, pages 2071–2080, 2016. 3 7 Domain and Range Aware Negatives for KGE

  27. [35]

    Canonical tensor decomposition for knowledge base completion

    Timothee Lacroix, Nicolas Usunier, and Guillaume Obozinski. Canonical tensor decomposition for knowledge base completion. In International Conference on Machine Learning , pages 2869–2878, 2018. 3

  28. [36]

    Tucker: Tensor factorization for knowledge graph completion

    Ivana Balaževi´c, Carl Allen, and Timothy M Hospedales. Tucker: Tensor factorization for knowledge graph completion. In Empirical Methods in Natural Language Processing, 2019. 3

  29. [37]

    A*net: A scalable path-based reasoning approach for knowledge graphs

    Zhaocheng Zhu, Xinyu Yuan, Mikhail Galkin, Sophie Xhonneux, Ming Zhang, Maxime Gazeau, and Jian Tang. A*net: A scalable path-based reasoning approach for knowledge graphs. arXiv preprint arXiv:2206.04798, 2022. 3

  30. [38]

    AmpliGraph: a Library for Representation Learning on Knowledge Graphs, March 2019

    Luca Costabello, Alberto Bernardi, Adrianna Janik, Aldan Creo, Sumit Pai, Chan Le Van, Rory McGrath, Nicholas McCarthy, and Pedro Tabacof. AmpliGraph: a Library for Representation Learning on Knowledge Graphs, March 2019. URL https://doi.org/10.5281/zenodo. 2595043. 4, 9 8 Dom...

  31. [962]

    doi: 10.18653/V1/P17-1088

    Association for Computational Linguistics, 2017. doi: 10.18653/V1/P17-1088. URL https://doi.org/10.18653/v1/P17-1088

  32. [2013]

    URL https://proceedings.neurips.cc/paper_files/paper/2013/file/ 1cecc7a77928ca8133fa24680a88d2f9-Paper.pdf. 1, 2, 3

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.