Hyperbolic Contrastive Learning with Model-augmentation for Knowledge-aware Recommendation

Chen Ma; Shengyin Sun

REVIEW 4 major objections 5 minor 38 references

Hyperbolic Contrastive Learning with Model-augmentation for Knowledge-aware Recommendation

T0 review · 4 major / 5 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 Replacing structural graph perturbation with model-level augmentation—dropout, cross-layer outputs, and pruning—and embedding in Lorentzian hyperbolic space improves knowledge-aware recommendation by up to 11.03%.

desk verdict A sensible integration paper whose SOTA claim needs multi-seed evidence and a hyperbolic baseline before the margins can be trusted. read the letter →

arxiv 2505.08157 v1 pith:2O6NW5IR submitted 2025-05-13 cs.IR cs.AI

Shengyin Sun , Chen Ma This is my paper · ORCID

classification cs.IRcs.AI

keywords knowledge-awarerecommendationhyperbolicrepresentationlearningLorentzmodelcontrastiveaugmentationknowledgegraphpreferenceshiftneuralnetworks

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper proposes HCMKR, a knowledge-aware recommender that learns user and item embeddings in the Lorentz model of hyperbolic space and builds contrastive positive pairs by perturbing the model rather than the graph. It aims to establish two things: hyperbolic aggregation captures the hierarchical, power-law structure of real interaction and knowledge graphs better than Euclidean encoders, and model-level augmentations—dropout masks, cross-layer outputs, and magnitude pruning—create positive views without shifting user preferences the way edge dropping does. On Yelp2018, Amazon-Book, and MovieLens-20M, the cross-layer variant HCMKR-C reports gains of 2.33% to 11.03% over the best baseline, with HCMKR-D close behind. If these results hold, structural graph augmentation is unnecessary for contrastive recommendation, and hyperbolic geometry plus model perturbation is a cheaper route to the same self-supervision.

What carries the argument

The load-bearing machinery is the Lorentz model of hyperbolic geometry, used through exponential and logarithmic maps so that aggregation happens in the tangent space while embeddings live on the manifold of negative curvature; this is supported by a relation-aware attention weight based on a Fermi-Dirac function of hyperbolic distance. The second piece is the family of model-level augmentations: dropout masks (two forward passes of the same input), cross-layer outputs (views from different GNN layers in a single pass), and magnitude pruning (zeroing weights under a threshold). Together they replace structural edge dropping, which the paper argues rewrites a user's purchase history and shifts preference semantics, with perturbations that keep the input graph and its supervision signal unchanged.

What would settle it

Measure preference drift directly: for a fixed user, compare the ranking of held-out items induced by the unaugmented representation with the ranking induced by each augmented view. If the edge-dropped view shifts the ranking toward irrelevant items no more than the model-augmented views do, the preference-shift premise fails.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that the two standard limitations of contrastive knowledge-aware recommendation can be addressed by the same design move: do the representation learning in the Lorentzian hyperbolic manifold, and generate contrastive views by altering the encoder rather than the input graph. The proposed Lorentzian knowledge aggregation propagates relation-weighted messages in the tangent space, then the hyperbolic encoder performs symmetric neighbor aggregation; the three model-level augmentations produce two views of every user and item without touching the user-item edges or knowledge-graph triplets. The reported experiments claim that HCMKR-C always ranks first across all metrics on all three datasets, with improvements from 2.33% to 11.03% over the strongest baseline, and that the learned embeddings are more uniformly distributed. The paper attributes the gain specifically to hierarchy capture from hyperbolic geometry and preference preservation from model-level augmentation.

Load-bearing premise

The key assumption is that randomly dropping neurons, using different layers' outputs, and pruning small weights create alternative versions of a user's representation that still reflect the same interests; the paper does not measure that directly.

Editorial extensions

If this is right

Structural edge dropping is not necessary for contrastive recommendation; model-level perturbations alone can supply the positive views.
A knowledge-aware recommender can run in hyperbolic space without a prohibitive time penalty: HCMKR-C trains at about 0.88 times the per-epoch cost of KGCL on the tested datasets.
Taking different GNN layer outputs as the two contrastive views is the strongest of the three augmentations, suggesting that depth itself provides a useful view separation.
The gains are largest on MovieLens-20M, the largest dataset here, so the benefit of hyperbolic modeling and model augmentation appears to scale with interaction volume.

Reading between the lines

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

If the preference-shift mechanism is real, the same model-augmentation recipe should transfer to graph contrastive tasks beyond recommendation, such as social or citation networks, where edge deletion also destroys node semantics.
A direct measurement of user-embedding drift under each augmentation would turn the Figure 2(e) preference-shift story into a quantitative claim.
Because the three augmentations differ in cost and effect, combining or adaptively selecting them per batch may outperform any single variant.
The per-epoch time comparison suggests efficiency, but an end-to-end comparison including tuning would be the more demanding test.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

The short version: this is a competent integration of existing ideas—hyperbolic (Lorentz) GNNs for knowledge-aware recommendation plus three model-level contrastive augmentations. It reports large gains on three standard benchmarks, and the math and tables are internally consistent. The soft spot is not the method but the evidence: single-run numbers, no error bars, no hyperbolic baselines. The 'always ranks first' claim is plausible but not yet airtight.

What's new: the specific combination of relation-aware Lorentzian knowledge aggregation and model-level augmentations (dropout, cross-layer outputs, magnitude pruning) to avoid preference shift. The components exist separately (HGCF, SimGRACE, XSimGCL, KGCL), but the paper's contribution is the integration, plus the cross-layer view in hyperbolic space. The code is available, which is a plus. The ablation (Fig. 5) shows both hyperbolic geometry and model-augmentation contribute, and the parameter sensitivity analysis is reasonable.

Soft spots, in order of importance. First, no seed-level variance or significance tests anywhere. The smallest margins in Table 2 are 0.26% (Yelp R@20 for HCMKR-P) and 0.49% (Amazon-Book R@10 for HCMKR-P); those are within the seed-to-seed noise I'd expect from contrastive GNN recommenders. HCMKR-C's gains are larger, but the paper treats all numbers as deterministic. Multi-seed reruns would fix this and should be the first request. Second, the paper argues hyperbolic geometry captures hierarchy but never compares against a hyperbolic recommender (HGCF and the fully hyperbolic GCN are cited but not run as baselines). That leaves the mechanism story under-supported. Third, the preference-shift argument in Section 3.4 is asserted from a cartoon rather than measured. It is plausible, and the results are consistent with it, but it isn't direct evidence.

That said, the architecture is reasonable and the paper is readable. A few self-citations (Refs. 11–13) look tangential, but nothing that changes the result.

Who this is for: people working on knowledge-aware recommendation or contrastive learning for GNNs. It deserves a serious referee; the empirical claim, if reproducible, is a new SOTA on three benchmarks. My recommendation: send to review, with a request for multi-seed results and at least one hyperbolic baseline.

Referee Report

4 major / 5 minor

Summary. The manuscript proposes HCMKR, a knowledge-aware recommender that represents users and items in Lorentzian hyperbolic space and applies three model-level augmentations (dropout, cross-layer outputs, magnitude pruning) to generate positive pairs for contrastive learning. The central claim is that, unlike structure-level augmentations such as edge dropping, these model-level augmentations avoid preference shift, and that the hyperbolic aggregation captures the hierarchical structure of user-item and knowledge graphs. The paper reports experiments on Yelp2018, Amazon-Book, and MovieLens-20M, claiming that HCMKR-C always ranks first with gains of 2.33% to 11.03% over the best baseline, supported by ablations, parameter-sensitivity plots, and a runtime comparison.

Significance. If the empirical claims hold, the paper makes two useful contributions: it introduces model-level augmentation as a principled alternative to structural augmentation in knowledge-aware contrastive recommendation, and it demonstrates that Lorentzian aggregation can be integrated with standard contrastive objectives without prohibitive computational cost. The manuscript provides code, the Lorentzian mathematics (Eqs. 1-12) is standard and internally consistent, and the ablation study is well structured. However, the central empirical claim is currently under-supported: the experimental protocol reports no variance estimates, and the hyperbolic component is not benchmarked against existing hyperbolic recommenders.

major comments (4)

[Section 4.2, Table 2] The headline claim that HCMKR-C 'always ranks first' with gains of 2.33% to 11.03% is not supported by the reported protocol, which contains no number of seeds, no standard deviations, and no significance tests. Margins as small as 2.33% (NDCG@10 on Yelp2018) and 1.38% (NDCG@10 on Amazon-Book for HCMKR-P) are within the seed-to-seed variation typically observed in contrastive GNN recommenders, so multi-seed runs with variance and significance testing are required.
[Section 3.4, Fig. 2(e)] The central mechanism claim that model-level augmentations avoid preference shift is asserted and illustrated but never directly measured. The paper should include a controlled comparison between model-level and structure-level augmentations under the same hyperbolic encoder, or a quantitative proxy for preference semantics (e.g., the change in top-ranked items per user before and after augmentation), to substantiate the claimed advantage over KGCL-style structural augmentation.
[Section 4, Table 2] No hyperbolic baseline is included in the comparison; existing hyperbolic recommenders such as HGCF [23] and the Poincaré-based knowledge-aware method [17] are cited in related work but absent from Table 2. Without such baselines, the contribution of the Lorentzian aggregation is not externally anchored, and the reported gains could in principle be attributed to the contrastive regularization rather than to hyperbolic geometry.
[Section 3.4, Eq. (16)] There is an internal contradiction in the pruning augmentation: the text states 'set the elements in Θ greater than the ξ to 0', but the mask M_P^2 = 1(|Θ|>ξ) keeps elements greater than ξ and zeros smaller elements. Please correct the threshold direction and clarify the pruning-ratio definition, since this determines the actual behavior of HCMKR-P and is needed for reproducibility.

minor comments (5)

[Section 3.2, Eq. (3)] The logarithmic map in Eq. (3) appears garbled in the typesetting; please restate the normalized vector expression clearly.
[Section 3.3, after Eq. (12)] The text describes z_i^(k) as 'the representation of the item node u'; this should be item node i.
[Table 2] The HCMKR-P Improve row is N/A on MovieLens-20M even though several metrics improve; report the comparison against KGCL explicitly, and note that R@20 is actually lower than KGCL on that dataset.
[Contributions bullet and Table 3] The claim of 'reduced time consumption (×0.88)' applies only to HCMKR-C; Table 3 shows HCMKR-P and HCMKR-D are slower than KGCL (×1.28 and ×1.27). Please qualify the efficiency claim accordingly.
[Section 3.5, Eq. (19)] The denominator of the contrastive loss uses the set j∈{i+}∪V\{i}; please clarify whether this includes the positive sample itself and whether duplicate indices are excluded, to make the loss unambiguous.

Circularity Check

0 steps flagged · score 2.0 of 10

No load-bearing circularity: HCMKR's claimed ranking gains are external benchmark measurements, not quantities defined by the paper's equations or fitted parameters.

full rationale

The derivation chain is self-contained: standard Lorentzian exp/log maps, relation-aware attention aggregation, tangent-space GCN propagation, three model-level augmentation views, and BPR plus InfoNCE losses. None of these equations defines the reported Recall@K/NDCG@K values in terms of a fitted input; the Table 2 numbers are obtained by ranking held-out test items and comparing with external baselines. The Section 3.4 claim that model-level augmentation 'avoids preference shifts' is a design rationale tied to not perturbing the graph, not a computed result that reduces to its own input. The only self-citations ([17], [28], and signal-processing items [11]-[13]) appear in related-work and background lists, and none is load-bearing for the central empirical claim or used to forbid alternative methods. There is no uniqueness theorem imported from the authors' prior work and no fitted parameter renamed as a prediction. Concerns about single-run comparisons and absent significance testing in Section 4.2 are experimental-validity issues, not circularity.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The model introduces no new physical entities. The central results rest on standard hyperbolic geometry background plus two paper-specific assumptions: that model-level augmentation preserves preference semantics and that magnitude pruning preserves output semantics. Hyperparameters (curvature, Fermi-Dirac constants, temperature, contrastive weight, dropout/pruning ratios, layer pair) are tuned on validation, and per-dataset optimal values are only partially reported. These are not fitted to test-set metrics, so the circularity burden is low, but the untested augmentation-fidelity assumptions carry the main risk.

free parameters (6)

curvature c of Lorentz model = learned (trainable), initialization not specified
Sets the geometry of the hyperbolic space; a trainable variable following [2].
Fermi-Dirac c1, c2 = searched in {0.5, 1.0, 1.5}
Controls the monotonicity transform in Eq. (11); selected by validation on each dataset, values not reported.
contrastive temperature tau = searched in {0.2, 0.4, 0.6, 0.8, 1.0}
Standard InfoNCE temperature in Eq. (19); validation-tuned.
contrastive weight lambda = searched in {0.1, 0.2, 0.3, 0.4, 0.5}; optimal per Fig. 6(d) e.g., 0.5/0.5/0.2
Balances L1 and L2 in Eq. (20); tuned per dataset.
dropout/pruning ratio = searched in {0.1, 0.2, 0.3, 0.4, 0.5}
Controls view diversity for Dropout and pruning augmentations; sensitivity in Fig. 6(a,b).
cross-layer view pair (k1,k2) = best 1-3 on Yelp2018; not reported for other datasets
Design choice for which layer outputs form positive pairs in Eq. (14); sensitivity in Fig. 6(c).

assumptions (5)

standard math Lorentz model exponential/logarithmic maps and distance, Eqs. (1)-(3) and (10), are valid and numerically stable in the regime used.
Imported from hyperbolic geometry literature (Law et al. 2019; Chami et al. 2019); assumed without proof.
domain assumption The power-law degree distributions observed in Fig. 1 imply hierarchical structure that hyperbolic embeddings represent better than Euclidean embeddings (Bourgain's theorem via [16,22]).
Used in Section 1 to motivate Lorentzian aggregation; the empirical link between power-law degree and recommendation-relevant hierarchy is assumed.
ad hoc to paper Structural perturbations of the interaction graph shift user preferences, while model-level perturbations do not.
Core premise of Section 3.4 and the abstract; no experiment in the paper measures preference shift or semantic drift directly.
ad hoc to paper Magnitude pruning (Eq. 16) with a threshold derived from a pruning ratio preserves the semantic similarity of outputs to the unpruned model.
Heuristic used for view P in Eq. (15); assumed to generate valid positive pairs.
domain assumption Feature transformation and non-linear activation are unnecessary in recommendation GNNs and cause overfitting on sparse graphs.
Inherited from LightGCN [7] and used in Eq. (12).

how reviews work

0 comments

Cite this review

Pith. "Pith review of Hyperbolic Contrastive Learning with Model-augmentation for Knowledge-aware Recommendation." pith.science (2026). https://pith.science/paper/2O6NW5IR

@misc{pith2026250508157,
  author       = {Pith},
  title        = {Pith review of: Hyperbolic Contrastive Learning with Model-augmentation for Knowledge-aware Recommendation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2O6NW5IR}},
  note         = {Machine review of arXiv:2505.08157}
}

abstract

Benefiting from the effectiveness of graph neural networks (GNNs) and contrastive learning, GNN-based contrastive learning has become mainstream for knowledge-aware recommendation. However, most existing contrastive learning-based methods have difficulties in effectively capturing the underlying hierarchical structure within user-item bipartite graphs and knowledge graphs. Moreover, they commonly generate positive samples for contrastive learning by perturbing the graph structure, which may lead to a shift in user preference learning. To overcome these limitations, we propose hyperbolic contrastive learning with model-augmentation for knowledge-aware recommendation. To capture the intrinsic hierarchical graph structures, we first design a novel Lorentzian knowledge aggregation mechanism, which enables more effective representations of users and items. Then, we propose three model-level augmentation techniques to assist Hyperbolic contrastive learning. Different from the classical structure-level augmentation (e.g., edge dropping), the proposed model-augmentations can avoid preference shifts between the augmented positive pair. Finally, we conduct extensive experiments to demonstrate the superiority (maximum improvement of $11.03\%$) of proposed methods over existing baselines.

Figures

Figures reproduced from arXiv: 2505.08157 by the authors.

**Figure 2.** Illustrations of the model-level augmentations and preference shift. (a) [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** Visualizing the distribution of embeddings (Yelp2018). [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗

Figures from the paper (4 more)

**Figure 4.** Figure 4: Visualizing the distribution of embeddings (Amazon-Book). [PITH_FULL_IMAGE:figures/full_fig_p014_4.png]

**Figure 5.** Figure 5: Ablation experiments of the proposed methods. [PITH_FULL_IMAGE:figures/full_fig_p014_5.png]

**Figure 5.** Figure 5: In experiments without using the Hyperbolic geometry and the model [PITH_FULL_IMAGE:figures/full_fig_p015_5.png]

**Figure 6.** Figure 6: Parameter sensitivity of Dropout ratio, prune ratio, combination of different layers, and λ (Yelp2018) [PITH_FULL_IMAGE:figures/full_fig_p015_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

38 extracted references · 36 canonical work pages

[23]

In: Proc

Sun, J., Cheng, Z., Zuberi, S., Perez, F., Volkovs, M.: Hgcf: Hyperbolic graph convolution networks for collaborative filtering. In: Proc. Int. Conf. World Wide Web (2021)

work page 2021
[17]

In: AAAI (2021)

Ma, C., Ma, L., Zhang, Y., Wu, H., Liu, X., Coates, M.: Knowledge-enhanced top-k recommendation in poincaré ball. In: AAAI (2021)

work page 2021
[1]

Algorithms (2018)

Ai, Q., Azizi, V., Chen, X., Zhang, Y.: Learning heterogeneous knowledge base embeddings for explainable recommendation. Algorithms (2018)

work page 2018
[2]

In: Advances in Neural Information Processing Systems (2019)

Chami, I., Ying, Z., Ré, C., Leskovec, J.: Hyperbolic graph convolutional neural networks. In: Advances in Neural Information Processing Systems (2019)

work page 2019
[3]

In: Proc

Chen, M., Huang, C., Xia, L., Wei, W., Xu, Y., Luo, R.: Heterogeneous graph contrastive learning for recommendation. In: Proc. ACM Int. Conf. Web Search and Data Mining (2023)

work page 2023
[4]

In: Proc

Chen, Y., Yang, M., Zhang, Y., Zhao, M., Meng, Z., Hao, J., King, I.: Modeling scale-free graphs with hyperbolic geometry for knowledge-aware recommendation. In: Proc. ACM Int. Conf. Web Search and Data Mining (2022)

work page 2022
[5]

In: Annales de l’institut Henri Poincaré (1948)

Fréchet, M.: Les éléments aléatoires de nature quelconque dans un espace distancié. In: Annales de l’institut Henri Poincaré (1948)

work page 1948
[6]

ACM Trans

Gao, C., Zheng, Y., Li, N., Li, Y., Qin, Y., Piao, J., Quan, Y., Chang, J., Jin, D., He, X., Li, Y.: A survey of graph neural networks for recommender systems: Challenges, methods, and directions. ACM Trans. Recommender Systems (2023)

work page 2023

Show all 38 references

[7]

In: Proc

He, X., Deng, K., Wang, X., Li, Y., Zhang, Y., Wang, M.: Lightgcn: Simplifying and powering graph convolution network for recommendation. In: Proc. ACM Int. Conf. Research and Development in Information Retrieval (2020)

2020
[8]

In: Proc

He, X., Liao, L., Zhang, H., Nie, L., Hu, X., Chua, T.S.: Neural collaborative filter- ing. In: Proc. Int. Conf. World Wide Web (2017)

2017
[9]

In: Proc

Hu, B., Shi, C., Zhao, W.X., Yu, P.S.: Leveraging meta-path based context for top-n recommendation with a neural co-attention model. In: Proc. ACM Conf. Knowledge Discovery and Data Mining (2018)

2018
[10]

Physical Review E (2010)

Krioukov, D., Papadopoulos, F., Kitsak, M., Vahdat, A., Boguna, M.: Hyperbolic geometry of complex networks. Physical Review E (2010)

2010
[11]

Pattern Recognition (2023)

Sun, S., Liu, J., Sun, S.: Hyperspectral subpixel target detection based on interac- tion subspace model. Pattern Recognition (2023)

2023
[12]

Signal Processing (2021)

Sun, S., Liu, J., Liu, W., Jian, T.: Robust detection of distributed targets based on Rao test and Wald test. Signal Processing (2021)

2021
[13]

EUSIPCO (2020)

Sun, S., Liu, J., Liu, W.: Rao test With improved robustness for range-spread target detection. EUSIPCO (2020)

2020
[14]

In: Proc

Law, M., Liao, R., Snell, J., Zemel, R.: Lorentzian distance learning for hyperbolic representations. In: Proc. Int. Conf. Machine Learning (2019)

2019
[15]

Information Sciences (2024)

Li, J., Yang, C., Ye, G., Nguyen, Q.V.H.: Graph neural networks with deep mutual learning for designing multi-modal recommendation systems. Information Sciences (2024)

2024
[16]

Combinatorica (1995)

Linial, N., London, E., Rabinovich, Y.: The geometry of graphs and some of its algorithmic applications. Combinatorica (1995)

1995
[18]

In: Proc

Nickel, M., Kiela, D.: Learning continuous hierarchies in the lorentz model of hy- perbolic geometry. In: Proc. Int. Conf. Machine Learning (2018)

2018
[19]

In: Advances in Neural Information Processing Systems (2017)

Nickel, M., Kiela, D.: Poincaré embeddings for learning hierarchical representa- tions. In: Advances in Neural Information Processing Systems (2017)

2017
[20]

In: Proc

Qiu, R., Huang, Z., Yin, H., Wang, Z.: Contrastive learning for representation degeneration problem in sequential recommendation. In: Proc. ACM Int. Conf. Web Search and Data Mining (2022) 18 S. Sun et al

2022
[21]

In: arXiv:1205.2618 (2012)

Rendle, S., Freudenthaler, C., Gantner, Z., Schmidt-Thieme, L.: Bpr: Bayesian personalized ranking from implicit feedback. In: arXiv:1205.2618 (2012)

2012 arXiv
[22]

In: Proc

Sala, F., Sa, C.D., Gu, A., Ré, C.: Representation tradeoffs for hyperbolic embed- dings. In: Proc. Int. Conf. Machine Learning (2018)

2018
[24]

In: Proc

Velickovic, P., Cucurull, G., Casanova, A., Romero, A., Liò, P., Bengio, Y.: Graph attention networks. In: Proc. Int. Conf. Learning Representations (2018)

2018
[25]

In: Proc

Wang, H., Zhang, F., Wang, J., Zhao, M., Li, W., Xie, X., Guo, M.: Ripplenet: Propagating user preferences on the knowledge graph for recommender systems. In: Proc. ACM Int. Conf. Information and Knowledge Management (2018)

2018
[26]

In: Proc

Wang, H., Zhang, F., Zhang, M., Leskovec, J., Zhao, M., Li, W., Wang, Z.: Knowledge-aware graph neural networks with label smoothness regularization for recommender systems. In: Proc. ACM Conf. Knowledge Discovery and Data Mining (2019)

2019
[27]

In: Proc

Wang, H., Zhao, M., Xie, X., Li, W., Guo, M.: Knowledge graph convolutional networks for recommender systems. In: Proc. Int. Conf. World Wide Web (2019)

2019
[28]

In: arXiv:2311.14324 (2024)

Sun, S., Ren, Y., Chen, J., Ma, C.: Large language models as topological structure enhancers for text-attributed graphs. In: arXiv:2311.14324 (2024)

2024 arXiv
[29]

In: Proc

Wang, L., Hu, F., Wu, S., Wang, L.: Fully hyperbolic graph convolution network for recommendation. In: Proc. ACM Int. Conf. Information and Knowledge Man- agement (2021)

2021
[30]

In: Proc

Wang, X., He, X., Cao, Y., Liu, M., Chua, T.S.: Kgat: Knowledge graph attention network for recommendation. In: Proc. ACM Conf. Knowledge Discovery and Data Mining (2019)

2019
[31]

In: IEEE Trans

Wei, C., Hu, C., Wang, C.D., Huang, S.: Time-aware multibehavior contrastive learning for social recommendation. In: IEEE Trans. Knowledge and Data Engi- neering (2024)

2024
[32]

In: Proc

Wu,J.,Wang,X.,Feng,F.,He,X.,Chen,L.,Lian,J.,Xie,X.:Self-supervisedgraph learning for recommendation. In: Proc. ACM Int. Conf. Research and Development in Information Retrieval (2021)

2021
[33]

In: Proc

Xia, J., Wu, L., Chen, J., Hu, B., Li, S.Z.: Simgrace: A simple framework for graph contrastive learning without data augmentation. In: Proc. Int. Conf. World Wide Web (2022)

2022
[34]

In: Proc

Xuan, H., Liu, Y., Li, B., Yin, H.: Knowledge enhancement for contrastive multi- behavior recommendation. In: Proc. ACM Int. Conf. Web Search and Data Mining (2023)

2023
[35]

In: Proc

Yang, Y., Huang, C., Xia, L., Li, C.: Knowledge graph contrastive learning for rec- ommendation. In: Proc. ACM Int. Conf. Research and Development in Information Retrieval (2022)

2022
[36]

IEEE Trans

Yu, J., Xia, X., Chen, T., Cui, L., Hung, N.Q.V., Yin, H.: Xsimgcl: Towards ex- tremely simple graph contrastive learning for recommendation. IEEE Trans. Knowl- edge and Data Engineering (2023)

2023
[37]

In: Proc

Zhang, F., Yuan, N.J., Lian, D., Xie, X., Ma, W.Y.: Collaborative knowledge base embedding for recommender systems. In: Proc. ACM Conf. Knowledge Discovery and Data Mining (2016)

2016
[38]

In: Proc

Zou, D., Wei, W., Mao, X.L., Wang, Z., Qiu, M., Zhu, F., Cao, X.: Multi-level cross-view contrastive learning for knowledge-aware recommender system. In: Proc. ACM Int. Conf. Research and Development in Information Retrieval (2022)

2022

Pith tools

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

[1] [23]

In: Proc

Sun, J., Cheng, Z., Zuberi, S., Perez, F., Volkovs, M.: Hgcf: Hyperbolic graph convolution networks for collaborative filtering. In: Proc. Int. Conf. World Wide Web (2021)

work page 2021

[2] [17]

In: AAAI (2021)

Ma, C., Ma, L., Zhang, Y., Wu, H., Liu, X., Coates, M.: Knowledge-enhanced top-k recommendation in poincaré ball. In: AAAI (2021)

work page 2021

[3] [1]

Algorithms (2018)

Ai, Q., Azizi, V., Chen, X., Zhang, Y.: Learning heterogeneous knowledge base embeddings for explainable recommendation. Algorithms (2018)

work page 2018

[4] [2]

In: Advances in Neural Information Processing Systems (2019)

Chami, I., Ying, Z., Ré, C., Leskovec, J.: Hyperbolic graph convolutional neural networks. In: Advances in Neural Information Processing Systems (2019)

work page 2019

[5] [3]

In: Proc

Chen, M., Huang, C., Xia, L., Wei, W., Xu, Y., Luo, R.: Heterogeneous graph contrastive learning for recommendation. In: Proc. ACM Int. Conf. Web Search and Data Mining (2023)

work page 2023

[6] [4]

In: Proc

Chen, Y., Yang, M., Zhang, Y., Zhao, M., Meng, Z., Hao, J., King, I.: Modeling scale-free graphs with hyperbolic geometry for knowledge-aware recommendation. In: Proc. ACM Int. Conf. Web Search and Data Mining (2022)

work page 2022

[7] [5]

In: Annales de l’institut Henri Poincaré (1948)

Fréchet, M.: Les éléments aléatoires de nature quelconque dans un espace distancié. In: Annales de l’institut Henri Poincaré (1948)

work page 1948

[8] [6]

ACM Trans

Gao, C., Zheng, Y., Li, N., Li, Y., Qin, Y., Piao, J., Quan, Y., Chang, J., Jin, D., He, X., Li, Y.: A survey of graph neural networks for recommender systems: Challenges, methods, and directions. ACM Trans. Recommender Systems (2023)

work page 2023