Pith. sign in

REVIEW 5 major objections 6 minor 61 references

Complementarity-driven Representation Learning for Multi-modal Knowledge Graph Completion

T0 review · 5 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read MoCME claims that weighting each expert view and each modality by negative mutual information, then training with entropy-weighted hard negatives, produces state-of-the-art multi-modal knowledge graph completion on five benchmarks.

desk verdict The central SOTA claim is unverifiable as written because MoMoK—the direct predecessor MoCME is built on—is missing from every baseline table, so the paper needs major revision before its empirical headline can be trusted. read the letter →

arxiv 2507.20620 v1 pith:6L6NVUPN submitted 2025-07-28 cs.AI cs.CV

Lijian Li This is my paper
classification cs.AIcs.CV
keywords multi-modalknowledgegraphcompletionmodalitycomplementaritymixtureofexpertsmutualinformationentropy-guidednegativesamplingRotatElinkprediction
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 tries to establish that multimodal knowledge graph completion improves when fusion weights reflect complementarity rather than attention or gating: each modality and each expert view is weighted by how little information it shares with its peers. It packages this as MoCME, built from a complementarity-guided fusion module and an entropy-guided negative-sampling loss. A careful reader should care because existing methods treat modalities as additive evidence, while this work treats redundancy itself as the signal to down-weight, which is a different principle. The paper reports that MoCME reaches the best MRR, Hit@1, and Hit@10 among all compared methods on five benchmark datasets, with the largest margins on datasets carrying four modalities.

What carries the argument

The load-bearing object is the complementarity weight: for a set of views, the weight of view $i$ is $\exp(-\sum_{j\neq i} I(v^{(i)}; v^{(j)})) / \sum_{i'} \exp(-\sum_{j\neq i'} I(v^{(i')}; v^{(j)}))$, using the mutual information neural estimator (MINE) to measure overlap; the same operation is applied again across modalities. It converts an information-theoretic notion of redundancy into concrete fusion coefficients, and the paper's argument is that this conversion is what distinguishes MoCME from concatenation, gating, and attention. The second mechanism is entropy-based negative sampling, which partitions negative triples by the binary entropy $\mathcal{H} = -p\log p - (1-p)\log(1-p)$ of their predicted plausibility into easy, ambiguous, and hard buckets with weights satisfying $\lambda_{\text{easy}} < \lambda_{\text{amb}} < \lambda_{\text{hard}}$.

What would settle it

On a fixed dataset such as DB15K, rerun MoCME with the complementarity weights replaced by uniform random weights while keeping everything else identical; if MRR does not drop materially below 39.62, the complementarity signal is not what carries the reported result.

Watch

Extended reading notes

Core claim

The central claim is that entity representations become better for link prediction when each view and each modality is weighted by its non-redundancy, estimated as negative mutual information between distributions of expert outputs. Intra-modality, the framework fuses K expert views of one modality with softmax weights over negative pairwise mutual information; inter-modality, it fuses the resulting modality embeddings the same way. The fused embedding is scored with RotatE, and training uses a negative-sampling loss in which easy, ambiguous, and hard negatives receive three different weights derived from the binary entropy of each negative's score. The paper argues this hierarchy of complementarity weighting is what lets the model keep informative signals when modalities are missing or noisy, and the experiments are offered as evidence that the design pays off, especially on datasets with audio, video, text, and image modalities.

Load-bearing premise

The approach assumes that the mutual-information estimates between expert views and between modalities are stable and meaningful, so the fusion weights reward genuinely non-redundant information rather than noise; if those estimates are biased, the weights are effectively arbitrary.

Editorial extensions

If this is right

  • If the central claim holds, state-of-the-art scores on all five benchmarks shift to a method whose fusion weights come from information overlap rather than learned attention, with the largest reported gains on the four-modality datasets TIVA and KVC16K.
  • A direct corollary of the ablations is that each modality contributes, and that removing the inter-modality complementarity layer costs more than removing any single modality, so cross-modal redundancy is the main carrier of the improvement.
  • The entropy-guided negative sampler implies that the same model trains better when easy negatives are down-weighted relative to ambiguous and hard ones, making negative-sample difficulty a tunable component of multimodal knowledge graph completion.
  • On datasets with only image and text, the gains over the previous best are smaller, so the claimed advantage is conditional on modality richness.

Reading between the lines

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

  • An implication the paper leaves implicit: the complementarity-weighting recipe is not tied to RotatE, so the same two-level mutual-information fusion could be grafted onto other knowledge graph scoring functions and tested as a general plug-in.
  • A testable extension the paper does not run: freeze the trained expert views and replace MINE with a cheaper closed-form correlation estimate; if MRR survives, the specific estimator is not the active ingredient, and the method would scale to larger graphs.
  • The paper's framing suggests a curriculum reading of the entropy sampler: thresholds could be annealed during training rather than fixed, which the paper does not explore but its own sensitivity analysis makes plausible.
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

5 major / 6 minor

Summary. The manuscript proposes MoCME, a multi-modal knowledge graph completion framework combining a Complementarity-guided Modality Knowledge Fusion (CMKF) module and an Entropy-guided Negative Sampling (EGNS) mechanism. CMKF uses per-modality expert networks and mutual information estimates to weight multi-view and multi-modal embeddings, while EGNS reweights negative samples by their binary entropy. The paper reports experiments on MKG-W, MKG-Y, DB15K, KVC16K, and TIVA and claims consistent state-of-the-art results over 19 baselines.

Significance. If the claimed results hold, the paper would make a useful empirical contribution to MMKGC by showing that complementarity-based weighting and entropy-guided negative sampling improve over plain fusion. Strengths include the clear problem framing, the inclusion of five datasets with varying modality coverage, and ablations of components, fusion strategies, backbones, expert counts, and thresholds. However, the paper's central SOTA claim is currently not sufficiently supported: the most relevant prior model is omitted from the comparisons, key components are underspecified, and no variability information is reported. The contribution is therefore better assessed as promising but requiring substantial revision rather than as an established result.

major comments (5)
  1. [§3.2, §4.3, Table 2] The paper states in §3.2 that the expert design is 'inspired by MoMoK [54]', yet MoMoK is not included among the 19 baselines in Table 2. Because MoMoK is the direct predecessor using per-modality mixtures of knowledge experts for MMKGC, the central claim in §4.4 that MoCME 'consistently achieves the best overall performance' cannot be evaluated without this comparison. Please add MoMoK to the baseline tables and report its MRR, Hit@1, and Hit@10 on all five datasets.
  2. [§3.2, Eqs. (2)-(4)] The complementarity weights in Eqs. (3) and (4) depend on mutual information estimates obtained with MINE, but the manuscript never specifies the MINE architecture, its training procedure, or how estimates are aggregated across entities and batches. The citation [22] also does not appear to be the MINE paper. Since these weights are the core novelty of CMKF, the current description is not reproducible and the reported weights could be arbitrary. Please provide full details or replace the estimator with an explicitly defined, computable quantity.
  3. [§4.1, Table 1] For KVC16K and TIVA, which include audio and video modalities, the paper reports large gains over NativE, but no audio or video encoder is named anywhere in §4.1; only VGG16 and BERT are mentioned. Please specify the encoders, input representations, and preprocessing for audio and video, and confirm whether the same frozen-encoder pipeline applies to these modalities.
  4. [§3.2] The description of the structural modality is internally contradictory: the text first says structure is 'encoded into token sequences and encode them using BERT', and then states that 'the final structural embedding eS is not directly extracted from a frozen encoder but rather initialized as a learnable parameter'. Please clarify which mechanism is actually used, because the two formulations lead to different models and affect the interpretation of the structural-modality results.
  5. [§4.1, Table 2] All reported numbers are averages over three independent runs, but no standard deviations or significance tests are given. The headline gains over NativE on MKG-W and MKG-Y are only +1.21 and +1.33 MRR, which may be within run-to-run noise. Please report standard deviations and, where feasible, statistical significance of the differences.
minor comments (6)
  1. [§4.5.1] The text states that removing image features causes the largest MRR decrease (-0.57), but Table 3 shows text removal gives the largest decrease (-1.21). Please correct this contradiction.
  2. [§4.5.5] The threshold grid search is described as being reported in Figure 2, but the corresponding figure is Figure 3; the expert-count ablation is Figure 2.
  3. [§4.5.4] The claim that the optimal expert count tracks the modality count 'across all three datasets' is not supported, since only DB15K is described in that paragraph.
  4. [Eq. (3)] The notation ωa_m · ve,m = Σ_i ... is confusing because ωa_m is a vector of weights; please rewrite with explicit index notation.
  5. [References] The reference list contains incomplete entries (e.g., [22] has no venue or year, [55] is listed only as '(No Title)'); please check all references.
  6. [Reproducibility] Please state whether code will be released, as the method currently cannot be re-implemented from the text alone.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the SOTA claim rests on external benchmarks, and the mutual-information/entropy signals are internal training mechanisms rather than predictions derived from fitted outputs.

full rationale

No significant circularity found. The paper's central claims are empirical: MoCME is evaluated against external benchmarks (MKG-W, MKG-Y, DB15K, TIVA, KVC16K) and 19 baselines with results reported in Table 2. The complementarity weighting (Eqs. 2-4) computes mutual information between the model's own view/modality embeddings as an adaptive fusion signal; this is an internal training mechanism, not a prediction derived from fitted outputs, so it does not reduce Eq. X to Eq. Y by construction. The EGNS entropy weighting (Eqs. 6-8) uses the model's current scores to weight its own negatives; again this is a self-paced training signal, not a circular derivation of a claimed result. Hyperparameters (delta1, delta2, lambda) are tuned on validation, which is standard practice and not circular. The paper cites many prior works by the same author in the related-work section, but none of these citations is load-bearing for the main result; the core architecture is attributed to MoMoK [54], which is a non-self citation. The omission of MoMoK from Table 2 is an empirical/rigor concern (missing comparator) but not circularity. Accordingly, the derivation chain is self-contained with respect to the external evaluation, and the score is 0.

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

MoCME introduces no new physical or conceptual entities beyond internal expert views and learned weights. The framework relies on several domain assumptions: MINE-based complementarity is meaningful, entropy indicates negative difficulty, pretrained encoders cover all modalities, and the structural modality can be represented in a learnable embedding. The main free parameters are the entropy thresholds, loss weights, expert count, and standard training hyperparameters, all selected on validation data.

free parameters (5)
  • Entropy thresholds delta1, delta2 = delta1=0.2, delta2=0.8
    Chosen by grid search on validation MRR in Section 4.5.5. Used to partition negatives into easy, ambiguous, and hard categories in EGNS.
  • Negative-type loss weights lambda_easy, lambda_amb, lambda_hard = 0.5, 1.5, 1.2
    Set by hand in Section 4.1. No sensitivity analysis is reported for these weights.
  • Number of experts K = 3 for DB15K
    Selected from ablation in Section 4.5.4. The best value matches the number of modalities, but this interpretation is post hoc.
  • Embedding and projection dimension = 256
    Fixed across entities, relations, and modality projections, following NATIVE in Section 4.1.
  • Learning rate, batch size, epochs = 1e-4, 1024, 1000 with early stopping
    Adopted from NATIVE in Section 4.1; these choices affect all reported results.
assumptions (5)
  • domain assumption Mutual information between expert-view distributions, estimated with MINE, is a stable and meaningful measure of complementarity.
    Equations (2)-(4) use negative mutual information as softmax weights. The paper gives no training details for MINE, so the reliability of this assumption is untested.
  • domain assumption Binary entropy of the RotatE score probability is a valid measure of negative-sample difficulty and informativeness.
    Equations (6)-(7) partition negatives by entropy. This assumes the model's own probability p=sigma(S-) reflects semantic closeness to positives.
  • domain assumption Pre-trained VGG16 and BERT encoders provide useful features for all entity modalities, including numeric, audio, and video.
    Section 3.2 names VGG16 for visual and BERT for textual inputs, but does not state which encoders handle audio or video, or how numeric tokenization works with BERT.
  • ad hoc to paper The structural modality can be represented both as BERT-token sequences and as a learnable entity embedding eS.
    Section 3.2 introduces both representations but does not define how they interact. If the BERT structural tokens are built from full KG triples, test leakage is possible.
  • domain assumption The evaluation protocol, data splits, and negative sampling setup follow NATIVE and the WildKGC benchmarks without modification.
    Section 4.1 states hyperparameters follow NATIVE, but the exact split filtering and baseline tuning details are not fully described.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Complementarity-driven Representation Learning for Multi-modal Knowledge Graph Completion." pith.science (2026). https://pith.science/paper/6L6NVUPN

@misc{pith2026250720620,
  author       = {Pith},
  title        = {Pith review of: Complementarity-driven Representation Learning for Multi-modal Knowledge Graph Completion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6L6NVUPN}},
  note         = {Machine review of arXiv:2507.20620}
}
read the original abstract

Multi-modal Knowledge Graph Completion (MMKGC) aims to uncover hidden world knowledge in multimodal knowledge graphs by leveraging both multimodal and structural entity information. However, the inherent imbalance in multimodal knowledge graphs, where modality distributions vary across entities, poses challenges in utilizing additional modality data for robust entity representation. Existing MMKGC methods typically rely on attention or gate-based fusion mechanisms but overlook complementarity contained in multi-modal data. In this paper, we propose a novel framework named Mixture of Complementary Modality Experts (MoCME), which consists of a Complementarity-guided Modality Knowledge Fusion (CMKF) module and an Entropy-guided Negative Sampling (EGNS) mechanism. The CMKF module exploits both intra-modal and inter-modal complementarity to fuse multi-view and multi-modal embeddings, enhancing representations of entities. Additionally, we introduce an Entropy-guided Negative Sampling mechanism to dynamically prioritize informative and uncertain negative samples to enhance training effectiveness and model robustness. Extensive experiments on five benchmark datasets demonstrate that our MoCME achieves state-of-the-art performance, surpassing existing approaches.

Figures

Figures reproduced from arXiv: 2507.20620 by the authors.

Figure 1
Figure 1. The overview of the proposed MoCME framework, which consists of two main modules: a Complementary-guided Modality [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗
Figure 2
Figure 2. Ablation study on the number of experts in MoCME framework [PITH_FULL_IMAGE:figures/full_fig_p018_2.png] view at source ↗
Figure 3
Figure 3. MRR performance under different threshold settings of [PITH_FULL_IMAGE:figures/full_fig_p019_3.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

61 extracted references · 36 canonical work pages

  1. [54]

    Zhang, Z

    Y. Zhang, Z. Chen, L. Guo, Y. Xu, B. Hu, Z. Liu, W. Zhang, H. Chen, Mixture of modality knowledge experts for robust multi-modal knowledge graph completion, arXiv preprint arXiv:2405.16869 (2024)

    arXiv 2024

  2. [22]

    S. Li, C. Du, Y. Huang, L. Huang, H. Zhao, Modality complementariness: Towards understanding multi- modal robustness

    work page

  3. [1]

    Bordes, N

    A. Bordes, N. Usunier, A. Garcia-Duran, J. Weston, O. Yakhnenko, Translating embeddings for modeling multi-relational data, Advances in neural information processing systems 26 (2013)

    work page 2013

  4. [2]

    L. Chao, J. He, T. Wang, W. Chu, Pairre: Knowledge graph embeddings via paired relation vectors, arXiv preprint arXiv:2011.03798 (2020)

    arXiv 2020

  5. [3]

    Sun, Z.-H

    Z. Sun, Z.-H. Deng, J.-Y. Nie, J. Tang, Rotate: Knowledge graph embedding by relational rotation in complex space, arXiv preprint arXiv:1902.10197 (2019)

    arXiv 2019

  6. [4]

    Trouillon, J

    T. Trouillon, J. Welbl, S. Riedel, ´E. Gaussier, G. Bouchard, Complex embeddings for simple link prediction, in: International conference on machine learning, PMLR, 2016, pp. 2071–2080

    work page 2016

  7. [5]

    J. Hu, H. Yang, F. Teng, S. Du, T. Li, A knowledge graph completion model based on triple level interaction and contrastive learning, Pattern Recognition 156 (2024) 110783

    work page 2024

  8. [6]

    Z. Wang, L. Yang, Z. Lei, A. U. Haq, D. Zhang, S. Yang, A. O. Francis, An entity-weights-based convolutional neural network for large-sale complex knowledge embedding, Pattern Recognition 131 (2022) 108841. 20

    work page 2022

Show all 61 references
  1. [7]

    Zheng, L

    W. Zheng, L. Yin, X. Chen, Z. Ma, S. Liu, B. Yang, Knowledge base graph embedding module design for visual question answering model, Pattern recognition 120 (2021) 108153

    2021

  2. [8]

    Shang, Y

    C. Shang, Y. Tang, J. Huang, J. Bi, X. He, B. Zhou, End-to-end structure-aware convolutional networks for knowledge base completion, in: The Thirty-Third AAAI Conference on Artificial Intelligence, AAAI 2019, The Thirty-First Innovative Applications of Artificial Intelligence ...

    2019 doi

  3. [9]

    D. Q. Nguyen, T. D. Nguyen, D. Q. Nguyen, D. Q. Phung, A novel embedding model for knowledge base completion based on convolutional neural network, in: M. A. Walker, H. Ji, A. Stent (Eds.), Proceedings of the 2018 Conference of the North American Chapter of the Association for...

    2018 doi

  4. [10]

    Vashishth, S

    S. Vashishth, S. Sanyal, V . Nitin, P . P . Talukdar, Composition-based multi-relational graph convolutional net- works, in: 8th International Conference on Learning Representations, ICLR 2020, Addis Ababa, Ethiopia, April 26-30, 2020, OpenReview.net, 2020. URL https://openrev...

    2020

  5. [11]

    L. Cai, W. Y. Wang, Kbgan: Adversarial learning for knowledge graph embeddings, arXiv preprint arXiv:1711.04071 (2017)

    2017 arXiv

  6. [12]

    Zhang, M

    Y. Zhang, M. Chen, W. Zhang, Modality-aware negative sampling for multi-modal knowledge graph embed- ding, in: 2023 International Joint Conference on Neural Networks (IJCNN), IEEE, 2023, pp. 1–8

    2023

  7. [13]

    D. Xu, T. Xu, S. Wu, J. Zhou, E. Chen, Relation-enhanced negative sampling for multimodal knowledge graph completion, in: Proceedings of the 30th ACM international conference on multimedia, 2022, pp. 3857–3866

    2022

  8. [14]

    Zhang, W

    Y. Zhang, W. Zhang, Knowledge graph completion with pre-trained multimodal transformer and twins neg- ative sampling, arXiv preprint arXiv:2209.07084 (2022)

    2022 arXiv

  9. [15]

    Z. Cao, Q. Xu, Z. Yang, Y. He, X. Cao, Q. Huang, Otkge: Multi-modal knowledge graph embeddings via optimal transport, Advances in Neural Information Processing Systems 35 (2022) 39090–39102

    2022

  10. [16]

    X. Wang, X. He, Y. Cao, M. Liu, T.-S. Chua, Kgat: Knowledge graph attention network for recommendation, in: Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining, 2019, pp. 950–958

    2019

  11. [17]

    X. Li, X. Zhao, J. Xu, Y. Zhang, C. Xing, Imf: interactive multimodal fusion model for link prediction, in: Proceedings of the ACM Web Conference 2023, 2023, pp. 2572–2580. 21

    2023

  12. [18]

    J. Lee, C. Chung, H. Lee, S. Jo, J. Whang, Vista: Visual-textual knowledge graph representation learning, in: Findings of the Association for Computational Linguistics: EMNLP 2023, 2023, pp. 7314–7328

    2023

  13. [19]

    Zhang, Z

    Y. Zhang, Z. Chen, L. Guo, Y. Xu, B. Hu, Z. Liu, W. Zhang, H. Chen, Native: Multi-modal knowledge graph completion in the wild, in: Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval, 2024, pp. 91–101

    2024

  14. [20]

    R. Xie, Z. Liu, H. Luan, M. Sun, Image-embodied knowledge representation learning, in: C. Sierra (Ed.), Pro- ceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence, IJCAI 2017, Melbourne, Australia, August 19-25, 2017, ijcai.org, 2017, pp. 3140–...

    2017 doi

  15. [21]

    Zhang, S

    Y. Zhang, S. Qian, Q. Fang, C. Xu, Multi-modal knowledge-aware hierarchical attention network for explain- able medical question answering, in: L. Amsaleg, B. Huet, M. A. Larson, G. Gravier, H. Hung, C. Ngo, W. T. Ooi (Eds.), Proceedings of the 27th ACM International Conferenc...

    2019

  16. [23]

    Liang, L

    K. Liang, L. Meng, M. Liu, Y. Liu, W. Tu, S. Wang, S. Zhou, X. Liu, F. Sun, K. He, A survey of knowledge graph reasoning on graph types: Static, dynamic, and multi-modal, IEEE Transactions on Pattern Analysis and Machine Intelligence (2024)

    2024

  17. [24]

    Q. Wang, Z. Mao, B. Wang, L. Guo, Knowledge graph embedding: A survey of approaches and applications, IEEE transactions on knowledge and data engineering 29 (12) (2017) 2724–2743

    2017

  18. [25]

    Yang, W.-t

    B. Yang, W.-t. Yih, X. He, J. Gao, L. Deng, Embedding entities and relations for learning and inference in knowledge bases, arXiv preprint arXiv:1412.6575 (2014)

    2014 arXiv

  19. [26]

    Bala ˇzevi´c, C

    I. Bala ˇzevi´c, C. Allen, T. M. Hospedales, Tucker: Tensor factorization for knowledge graph completion, arXiv preprint arXiv:1901.09590 (2019)

    2019 arXiv

  20. [27]

    M. Wang, S. Wang, H. Yang, Z. Zhang, X. Chen, G. Qi, Is visual context really helpful for knowledge graph? a representation learning perspective, in: Proceedings of the 29th ACM International Conference on Multime- dia, 2021, pp. 2735–2743

    2021

  21. [28]

    R. Xie, Z. Liu, H. Luan, M. Sun, Image-embodied knowledge representation learning, arXiv preprint arXiv:1609.07028 (2016)

    2016 arXiv

  22. [29]

    Zhang, Y

    Q. Zhang, Y. Wei, Z. Han, H. Fu, X. Peng, C. Deng, Q. Hu, C. Xu, J. Wen, D. Hu, et al., Multimodal fusion on low-quality data: A comprehensive survey, arXiv preprint arXiv:2404.18947 (2024). 22

    2024 arXiv

  23. [30]

    Wanyan, X

    Y. Wanyan, X. Yang, C. Chen, C. Xu, Active exploration of multimodal complementarity for few-shot action recognition, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 6492–6502

    2023

  24. [31]

    Y. He, F. Xiao, A new base function in basic probability assignment for conflict management, Applied Intelli- gence 52 (4) (2022) 4473–4487

    2022

  25. [32]

    Y. He, F. Xiao, Conflicting management of evidence combination from the point of improvement of basic probability assignment, International Journal of Intelligent Systems 36 (5) (2021) 1914–1942

    2021

  26. [33]

    Y. He, Y. Deng, Mmget: a markov model for generalized evidence theory, Computational and Applied Math- ematics 41 (2022) 1–41

    2022

  27. [34]

    Y. He, Y. Deng, Tdqmf: Two-dimensional quantum mass function, Information Sciences 621 (2023) 749–765

    2023

  28. [35]

    Y. He, Y. Deng, Ordinal belief entropy, Soft Computing 27 (11) (2023) 6973–6981

    2023

  29. [36]

    Y. He, Y. Deng, Ordinal fuzzy entropy, Iranian Journal of Fuzzy Systems 19 (3) (2022) 171–186

    2022

  30. [37]

    He, Epl: Evidential prototype learning for semi-supervised medical image segmentation, arXiv preprint arXiv:2404.06181 (2024)

    Y. He, Epl: Evidential prototype learning for semi-supervised medical image segmentation, arXiv preprint arXiv:2404.06181 (2024)

    2024 arXiv

  31. [38]

    Y. He, Y. Bi, L. Li, C.-M. Pun, W. Jiao, Z. Jin, Mutual evidential deep learning for semi-supervised medical im- age segmentation, in: 2024 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), IEEE, 2024, pp. 2010–2017

    2024

  32. [39]

    Y. He, L. Li, Uncertainty-aware evidential fusion-based learning for semi-supervised medical image segmen- tation, arXiv preprint arXiv:2404.06177 (2024)

    2024 arXiv

  33. [40]

    Y. He, L. Li, T. Zhan, C.-M. Pun, W. Jiao, Z. Jin, Co-evidential fusion with information volume for semi- supervised medical image segmentation, Pattern Recognition 166 (2025) 111639

    2025

  34. [41]

    L. Li, Y. He, C.-M. Pun, Efficient prototype consistency learning in semi-supervised medical image segmen- tation via joint uncertainty and data augmentation, in: 2024 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), IEEE, 2024, pp. 2114–2121

    2024

  35. [42]

    Y. He, L. Li, T. Zhan, A matrix-based distance of pythagorean fuzzy set and its application in medical diag- nosis, arXiv preprint arXiv:2102.01538 (2021)

    2021 arXiv

  36. [43]

    L. Li, Y. He, C.-M. Pun, An adaptive framework for multi-view clustering leveraging conditional entropy op- timization, in: ICASSP 2025-2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2025

    2025

  37. [44]

    Huang, Y

    C.-j. Huang, Y. He, X. Han, W. Jiao, Z. Jin, L. Wang, Unitrans: A unified vertical federated knowledge transfer framework for enhancing cross-hospital collaboration, arXiv preprint arXiv:2501.11388 (2025). 23

    2025 arXiv

  38. [45]

    Y. Bi, E. Che, Y. Chen, Y. He, J. Qu, Multi-prototype-based embedding refinement for medical image seg- mentation, in: ICASSP 2025-2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), IEEE, 2025, pp. 1–5

    2025

  39. [46]

    L. Li, Z. Jin, X. Zhang, H. Duan, J. Wang, Z. Tao, H. Zhao, X. Zhu, Multi-view riemannian manifolds fu- sion enhancement for knowledge graph completion, IEEE Transactions on Knowledge and Data Engineering (2025)

    2025

  40. [47]

    L. Li, Y. He, L. Li, Nndf: A new neural detection network for aspect-category sentiment analysis, in: Interna- tional Conference on Knowledge Science, Engineering and Management, Springer, 2022, pp. 339–355

    2022

  41. [48]

    Y. He, L. Li, T. Zhan, W. Jiao, C.-M. Pun, Generalized uncertainty-based evidential fusion with hybrid multi- head attention for weak-supervised temporal action localization, in: ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)...

    2024

  42. [49]

    T. Xu, K. Yan, Y. He, S. Gao, K. Yang, J. Wang, J. Liu, Z. Liu, Spatio-temporal variability analysis of vege- tation dynamics in china from 2000 to 2022 based on leaf area index: A multi-temporal image classification perspective, Remote Sensing 15 (12) (2023) 2975

    2023

  43. [50]

    L. Li, Z. Jin, Y. He, D. Jin, H. Duan, Z. Tao, X. Zhang, J. Li, Rethinking regularization methods for knowledge graph completion, arXiv preprint arXiv:2505.23442 (2025)

    2025 arXiv

  44. [51]

    L. Li, Z. Jin, Y. Zhang, D. Jin, C. Dou, Y. He, X. Zhang, H. Zhao, Towards structure-aware model for multi- modal knowledge graph completion, arXiv preprint arXiv:2505.21973 (2025)

    2025 arXiv

  45. [52]

    X. Chen, Z. Tao, K. Zhang, C. Zhou, W. Gu, Y. He, M. Zhang, X. Cai, H. Zhao, Z. Jin, Revisit self-debugging with self-generated tests for code generation, arXiv preprint arXiv:2501.12793 (2025)

    2025 arXiv

  46. [53]

    Simonyan, A

    K. Simonyan, A. Zisserman, Very deep convolutional networks for large-scale image recognition, arXiv preprint arXiv:1409.1556 (2014)

    2014 arXiv

  47. [55]

    P . K. Diederik, Adam: A method for stochastic optimization, (No Title) (2014)

    2014

  48. [56]

    Y. Liu, H. Li, A. Garcia-Duran, M. Niepert, D. Onoro-Rubio, D. S. Rosenblum, Mmkg: multi-modal knowledge graphs, in: The Semantic Web: 16th International Conference, ESWC 2019, Portoroˇz, Slovenia, June 2–6, 2019, Proceedings 16, Springer, 2019, pp. 459–474

    2019

  49. [57]

    X. Wang, B. Meng, H. Chen, Y. Meng, K. Lv, W. Zhu, Tiva-kg: A multimodal knowledge graph with text, image, video and audio, in: Proceedings of the 31st ACM International Conference on Multimedia, 2023, pp. 2391–2399

    2023

  50. [58]

    H. Pan, Z. Zhai, Y. Zhang, R. Fu, M. Liu, Y. Song, Z. Wang, B. Qin, Kuaipedia: a large-scale multi-modal short-video encyclopedia, arXiv preprint arXiv:2211.00732 (2022). 24

    2022 arXiv

  51. [59]

    Mousselly-Sergieh, T

    H. Mousselly-Sergieh, T. Botschen, I. Gurevych, S. Roth, A multimodal translation-based approach for knowl- edge graph representation learning, in: Proceedings of the Seventh Joint Conference on Lexical and Compu- tational Semantics, 2018, pp. 225–234

    2018

  52. [60]

    Z. Wang, L. Li, Q. Li, D. Zeng, Multimodal data enhanced representation learning for knowledge graphs, in: 2019 International Joint Conference on Neural Networks (IJCNN), IEEE, 2019, pp. 1–8

    2019

  53. [61]

    X. Lu, L. Wang, Z. Jiang, S. He, S. Liu, Mmkrl: A robust embedding approach for multi-modal knowledge graph representation learning, Applied Intelligence (2022) 1–18. 25

    2022

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.