Pith. sign in

REVIEW 4 major objections 5 minor 61 references

TPCH: Tensor-interacted Projection and Cooperative Hashing for Multi-view Clustering

T0 review · 4 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read This paper tries to establish that enforcing an enhanced tensor nuclear norm on both the stacked projection matrices and the stacked hash codes produces more compact and distinguishable binary codes, leading to state-of-the-art clustering…

desk verdict A useful dual-tensor hashing idea with public code, but the printed optimization equations don't minimize the stated objective, so the paper's accuracy claims aren't yet supported as written. read the letter →

arxiv 2412.18847 v1 pith:KHLLCX3J submitted 2024-12-25 cs.LG

Zhongwen Wang , Xingfeng Li , Yinghui Sun , Quansen Sun , Yuan Sun , Han Ling , Jian Dai , Zhenwen Ren This is my paper · ORCID
classification cs.LG
keywords multi-viewclusteringbinaryhashingtensornuclearnormt-SVDlow-rankbipartitegraphlarge-scaleHammingspace
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that enforcing an enhanced tensor nuclear norm on both the stacked projection matrices and the stacked hash codes produces binary codes that are more compact, more distinguishable, and more robust to noise, and that this leads to state-of-the-art clustering accuracy on five large-scale multi-view datasets while keeping runtime linear in the number of samples. The core move is to treat the multiple view projections as slices of a third-order tensor and to apply a t-SVD-based low-rank constraint not only to the hash-code tensor but also to the projection tensor, so that views communicate during projection itself. A reader should care because multi-view clustering at large scale is bottlenecked by the quality of low-dimensional codes, and the paper reports a large gain over prior hash-based methods, for example ACC 0.70 versus 0.29 on Cifar-10, with a roughly 7 to 10 times speedup over the strongest competitor.

What carries the argument

The central object is the enhanced tensor nuclear norm (ETNN), defined on a third-order tensor $T$ as $\|T\|_{\mathrm{etnn}} = \|\bar{\mathcal{S}}\|_* + \zeta \|\mathcal{U} * \mathcal{B}^{-1}(\bar{\mathcal{S}}) * \mathcal{V}\|_*$, where $\bar{\mathcal{S}}$ is a low-rank approximation of the core tensor from t-SVD, $\mathcal{B}$ and $\mathcal{B}^{-1}$ are linear transforms between core tensors, and $\|\cdot\|_*$ is the nuclear norm of the unfolded tensor. This norm is applied to both the stacked projection tensor $\mathcal{Q}$ and the stacked hash tensor $\mathcal{B}$ in the objective (Eq. 2), coupling the per-view projections $Q_p^\top \phi(X_p)$ to the per-view binary codes $B_p$; an alternating-direction solver updates $Q_p$, $B_p$ (via a sign function), and the two tensors (via closed-form proximal steps whose theorem is deferred to supplementary). The mechanism is that rank reduction on the core tensor suppresses noise and redundant information in both spaces while the $\mathcal{B}/\mathcal{B}^{-1}$ transforms keep the low-rank structure aligned with the t-SVD geometry, yielding hash codes that are more compact and more distinguishable in Hamming space.

What would settle it

Run TPCH on a dataset where the cluster-discriminative information is deliberately placed in the high-rank components of the cross-view tensor, for example a synthetic multi-view set where class structure lives in the singular tubes that ETNN truncates; if clustering accuracy does not drop sharply relative to a variant that uses the full core tensor, the low-rank-core assumption is not doing the work claimed, and if it does drop, the assumption is confirmed.

Watch

Extended reading notes

Core claim

The paper's central claim is that the proposed TPCH method learns more compact and distinguishable hash codes by stacking view-specific projection matrices $\{Q_p\}$ into a tensor $\mathcal{Q}$ and view-specific hash matrices $\{B_p\}$ into a tensor $\mathcal{B}$, and then minimizing the enhanced tensor nuclear norm (ETNN) on both tensors. ETNN differs from the ordinary tensor nuclear norm by first taking a low-rank approximation of the t-SVD core tensor $\bar{\mathcal{S}}$ and then applying a pair of transforms $\mathcal{B}$ and $\mathcal{B}^{-1}$ that map between the core tensor and its transformed counterpart; the sum of the nuclear norms of these two pieces is what is penalized. The paper argues that this removes noise and redundant information while preserving the higher-order cross-view synergies that the standard tensor nuclear norm leaves unresolved, and that the resulting binary codes, averaged across views and clustered in Hamming space, yield state-of-the-art ACC, NMI, Purity, F-score, and ARI on SUNRGBD, Cifar-10, Caltech101, Caltech256, and 100 leaves, with a time complexity linear in the sample count $n$.

Load-bearing premise

The load-bearing premise is that the low-rank approximation of the t-SVD core tensor, together with the $\mathcal{B}/\mathcal{B}^{-1}$ transforms, removes noise and redundancy without discarding the cross-view structure that separates clusters; if rank truncation cuts away class-discriminative signal, the claimed improvement collapses.

Editorial extensions

If this is right

  • Multi-view clustering on million-scale datasets becomes practical with substantially better accuracy than prior anchor or hash methods, since TPCH's cost is linear in the sample count.
  • The dual-tensor constraint yields binary codes that separate clusters more cleanly in Hamming space, so downstream tasks such as nearest-neighbor retrieval or clustering can use cheaper Hamming distances without losing accuracy.
  • TPCH remains stable under salt-and-pepper noise where the AC-MVBC baseline degrades, suggesting that core-tensor rank reduction is a robustness mechanism usable in other tensor factorization pipelines.
  • The reported 7 to 10 times CPU-time speedup over the GCAE baseline, if it holds across implementations, removes a practical barrier to large-scale multi-view analysis.

Reading between the lines

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

  • A testable extension is to apply the same ETNN dual-tensor trick to other paired-factor models, such as non-negative matrix factorization or dictionary learning over multiple views, where both the dictionary and the codes could be stacked and low-ranked.
  • The speed advantage hinges on the anchor count $m$ and code length $l$ staying small relative to $n$; on datasets where class structure forces $m$ to grow with $n$, the linear-time claim would erode, so the method's scalability is conditional on anchor sufficiency.
  • The paper's ablation removes $\mathcal{Q}$ and $\mathcal{B}$ separately, but it does not isolate whether the gain comes from core-tensor rank reduction or from the $\mathcal{B}/\mathcal{B}^{-1}$ transform; a variant with only the ordinary tensor nuclear norm on both tensors would pinpoint which piece drives the improvement.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes TPCH, an anchor- and hash-based multi-view clustering method. It stacks view-specific projection matrices Q_p into a tensor Q and the learned hash matrices B_p into a tensor B, and regularizes both tensors with an 'enhanced tensor nuclear norm' (ETNN) that penalizes a low-rank approximation of the t-SVD core tensor. An ADMM-type algorithm is presented that alternately updates Q_p, B_p, and auxiliary variables A and E; the final hash code is the average of the per-view codes and is clustered by discrete proximal linearized minimization. Experiments on five datasets report large accuracy gains over existing graph- and hash-based methods, together with linear-time complexity and public code.

Significance. If the algorithm as written actually minimized the objective in Eq. (2), the paper would make a useful contribution: the dual tensor low-rank idea on both projection matrices and hash codes is sensible, the linear-time complexity claim is attractive for large-scale multi-view clustering, and the public code is a concrete strength. However, the optimization derivations contain load-bearing errors that prevent the reader from verifying that Algorithm 1 optimizes the stated objective, and the key proximal update is deferred to a supplementary file that is not included. The empirical results are promising but are reported without variance information. The central claim is therefore not currently substantiated, although it appears fixable within the scope of a revision.

major comments (4)
  1. [Eq. (5)] The closed-form Q_p update printed immediately after Eq. (5) is not the stationary point of the displayed subproblem. With φ(X_p) ∈ R^{m×n}, Q_p ∈ R^{m×l}, and B_p ∈ R^{l×n}, the first-order condition for minimizing α||Q_p^T φ(X_p) - B_p||_F^2 + (μ/2)||Q_p - A_p + Y_p/μ||_F^2 with respect to Q_p is (2αφ(X_p)φ(X_p)^T + μI)Q_p = 2αφ(X_p)B_p^T + μA_p - Y_p, so the correct update is Q_p = (2αφ(X_p)φ(X_p)^T + μI)^{-1}(2αφ(X_p)B_p^T + μA_p - Y_p). The paper instead prints Q_p = [2αB_pφ^T(X_p) + μ(A_p - Y/μ)] / [2αφ(X_p)φ^T(X_p) + μI], whose numerator is l×m rather than m×l and which divides by a matrix on the right without specifying an inverse. As written, Algorithm 1's step 2 does not solve the Q_p subproblem of Eq. (5), so the algorithm is not shown to minimize Eq. (2). This is a load-bearing error and must be corrected.
  2. [Eq. (8)] The B_p update also contains a transpose error. In Eq. (6), the data term is ||Q_p^T φ(X_p) - B_p||_F^2; expanding this term and dropping the constant binary-norm terms gives a sgn argument of αQ_p^T φ(X_p) + (μ/2)(E_p - J_p/μ), which has dimension l×n and matches B_p. Equation (8) instead prints αQ_pφ(X_p) + (μ/2)(E_p - J_p/μ), which has dimension m×n under the stated dimensions and therefore cannot be an l×n hash matrix. Since this step is the discrete binarization of the hash codes, it is central to the method. Please correct the transpose and present the derivation explicitly.
  3. [Update-3/Update-4] The enhanced tensor nuclear norm is not sufficiently specified. Equation (3) does not define the low-rank approximation S̄ of the core tensor, the transformation pair B and B^{-1}, or the exact tensor dimensions (the text 'D ×D3' appears to describe a 2-D object rather than a 3-D tensor). More importantly, Update-3 states that the proximal step of A 'can produce a closed-form solution via Theorem 1 provided in supplementary materials,' and Update-4 is dismissed as similar, but the preprint contains no supplementary file and Eq. (9), which Algorithm 1 references for updating A, is missing from the manuscript. Consequently, Steps 4 and 5 of Algorithm 1 are not reproducible from the paper, and the claim that the ETNN-regularized objective is actually being optimized is unverified.
  4. [Table 1] The main empirical claim rests on unreplicated point estimates. Table 1 reports one ACC/NMI/Purity value per method per dataset with no standard deviations, no number of trials, and no significance tests. Given the exceptionally large reported margins on Cifar-10 (ACC 0.70 vs next-best 0.29, NMI 0.67 vs next-best 0.13), the absence of variance information makes it impossible to assess whether the improvement is stable or an artifact of a single run or of test-set hyperparameter tuning. Please report means and standard deviations over multiple runs and clarify the hyperparameter selection protocol (e.g., validation-based selection versus selection on the test metric).
minor comments (5)
  1. [Ablation Studies] The ablation paragraph says the methods without Q and without B are named 'without (w/o) B and without (w/o) B'; the second name should be 'without (w/o) Q', consistent with the legends in Fig. 4.
  2. [Fig. 6 caption] The caption contains the typo 'Synthtic 3d'; it should read 'Synthetic 3d'.
  3. [Table 1] The header of Table 1 has duplicated labels ('Acc Acc', 'NMI NMI') and the column alignment is broken; also, the F-score and ARI results are said to be in the supplementary materials, which are not included in the preprint.
  4. [Algorithm 1] The stopping criterion 'until Satisfy convergence' is vague; please specify a tolerance on the relative change of the objective in Eq. (2) or on the iterates.
  5. [Eq. (2)] The symbol B is used both for the stacked tensor B = Φ([B1; ...; Bv]) and for the binarized matrix constraint B ∈ {-1,1}^{l×n}; the paper should distinguish the tensor, the per-view matrices B_p, and the final averaged hash matrix B̂ defined later.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: TPCH's objective, updates, and empirical evaluation are self-contained; reported gains are not encoded in the objective.

full rationale

The derivation chain in TPCH is not circular. The training objective in Eq. (2) minimizes reconstruction error between projected bipartite graphs and binary codes plus tensor nuclear norms on the stacked projection and hash tensors; it contains no clustering labels, ACC, NMI, or other evaluation metrics that could be re-derived from the objective. The ETNN regularizer in Eq. (3) is a fixed low-rank penalty defined on t-SVD core tensors, not a quantity fitted to the target results. The alternating updates in Eqs. (5)-(12) are derived from the augmented Lagrangian of Eq. (4), and the final clustering step in Eq. (13) is applied after training to the learned hash codes; this is a post-processing evaluation, not a term that shapes the training loss. The paper's ablation and comparison experiments are external empirical checks. Self-citations to Li, Ren, Sun and co-workers appear only in the related-work survey and do not carry the derivation; the load-bearing cited tools (t-SVD/TNN solution via Lu et al. 2019 and DPLM via Wang et al. 2023) are independent of the present authors. The known weaknesses of the paper are correctness and reproducibility issues, not circularity: Eq. (5) prints a dimensionally inconsistent closed form for Q_p (left-multiplication versus right-division mismatch), Eq. (8) omits the transpose that Eq. (2) would require, and Theorem 1 for the proximal ETNN update is deferred to a missing supplementary proof. Those issues mean the algorithm as printed may not minimize Eq. (2), but they do not make any claimed prediction equivalent to its input by construction. Hyperparameter tuning of alpha on the test metric is selection bias, which is explicitly outside the circularity definition used here. No equation in the paper reduces to its own inputs, so the circularity score is 0.

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

The central contribution is an objective plus an optimization recipe, so the ledger is dominated by user-chosen hyperparameters and unproved modeling assumptions. No new physical entities are introduced. The main invented mathematical construct is the enhanced tensor nuclear norm, which is not independently validated and is defined through an opaque transformation pair.

free parameters (6)
  • alpha = not reported; searched over 1e-8 to 100
    Balance term between the projection error and the tensor regularizers; selected per dataset by logarithmic search on ACC, so the reported results are the best over a grid.
  • zeta = not reported
    Weight on the second term of the enhanced tensor nuclear norm in Eq. (3); called predefined but no value or tuning procedure is given.
  • kernel width delta = not reported
    Width of the RBF kernel used to build the bipartite graphs in Eq. (1); no selection rule or value is provided.
  • number of anchors m = not reported
    Anchors are randomly selected per view, but the number of anchors is never specified, despite it controlling approximation quality and complexity.
  • hash code length l = 64
    The dimensionality of the Hamming space is set to 64 by hand; no sensitivity analysis over this choice is reported.
  • ADMM penalty schedule = initial eta 1e-4, mu_max 1e10, rho=2
    Penalty parameters for the alternating optimization are chosen by hand and treated as implementation details.
assumptions (5)
  • standard math t-SVD and tensor nuclear norm have the singular-value properties used in the proximal updates.
    Imported from Lu et al. 2019 and used without proof in Eq. (3) and Update-3.
  • ad hoc to paper The transformation pair B and B^{-1} on the core tensor in ETNN is well-defined for all inputs.
    Defined only by Fig. 2 and Eq. (3); no formal construction or invertibility argument is given.
  • domain assumption Low-rank approximation of the core tensor removes noise while preserving class-discriminative cross-view structure.
    Motivates the ETNN in Eq. (3), but the paper offers no theorem or controlled experiment isolating this property.
  • domain assumption Randomly selected anchors preserve the original data structure after nonlinear RBF mapping.
    Stated near Eq. (1); no error bound or empirical validation is provided.
  • domain assumption Alternating minimization converges to a useful stationary point.
    Only empirical convergence curves are shown in Fig. 7(b); no convergence proof is given.

how reviews work

0 comments
Cite this review

Pith. "Pith review of TPCH: Tensor-interacted Projection and Cooperative Hashing for Multi-view Clustering." pith.science (2026). https://pith.science/paper/KHLLCX3J

@misc{pith2026241218847,
  author       = {Pith},
  title        = {Pith review of: TPCH: Tensor-interacted Projection and Cooperative Hashing for Multi-view Clustering},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KHLLCX3J}},
  note         = {Machine review of arXiv:2412.18847}
}
read the original abstract

In recent years, anchor and hash-based multi-view clustering methods have gained attention for their efficiency and simplicity in handling large-scale data. However, existing methods often overlook the interactions among multi-view data and higher-order cooperative relationships during projection, negatively impacting the quality of hash representation in low-dimensional spaces, clustering performance, and sensitivity to noise. To address this issue, we propose a novel approach named Tensor-Interacted Projection and Cooperative Hashing for Multi-View Clustering(TPCH). TPCH stacks multiple projection matrices into a tensor, taking into account the synergies and communications during the projection process. By capturing higher-order multi-view information through dual projection and Hamming space, TPCH employs an enhanced tensor nuclear norm to learn more compact and distinguishable hash representations, promoting communication within and between views. Experimental results demonstrate that this refined method significantly outperforms state-of-the-art methods in clustering on five large-scale multi-view datasets. Moreover, in terms of CPU time, TPCH achieves substantial acceleration compared to the most advanced current methods. The code is available at \textcolor{red}{\url{https://github.com/jankin-wang/TPCH}}.

Figures

Figures reproduced from arXiv: 2412.18847 by the authors.

Figure 1
Figure 1. In this way, TPCH captures the high-order semantic information among bipartite graphs and then embeds them into hash codes Bp ∈ {−1, 1} to enhance expression ability. From Eq. 2, TPCH could stack projection matrices and hash matrices to construct tensors Q and B to capture higher-order information of multi-view data in both dual projection and Hamming spaces. However, the tensors Q and B not only contain the desired… view at source ↗
Figure 2
Figure 2. Transformation instructions for core tensors. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 4
Figure 4. Ablation Studies of TPCH on the Caltech101 and [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figures from the paper (2 more)
Figure 5
Figure 5. Figure 5: Visualization of clustering results for TPCH and [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Clustering performance of TPCH at the beginning [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

61 extracted references · 57 canonical work pages

  1. [1]

    Chen, M.-S.; Wang, C.-D.; and Lai, J.-H. 2022. Low-rank tensor based proximity learning for multi-view clustering. IEEE Transactions on Knowledge and Data Engineering, 35(5): 5076--5090

    work page 2022

  2. [2]

    Chen, Z.; Wu, Z.; Wang, S.; and Guo, W. 2023. Dual low-rank graph autoencoder for semantic and topological networks. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, 4191--4198

    work page 2023

  3. [3]

    Fang, U.; Li, M.; Li, J.; Gao, L.; Jia, T.; and Zhang, Y. 2023. A Comprehensive Survey on Multi - View Clustering . IEEE Transactions on Knowledge and Data Engineering, 35(12): 12350--12368

    work page 2023

  4. [4]

    Gong, Y.; Lazebnik, S.; Gordo, A.; and Perronnin, F. 2012. Iterative quantization: A procrustean approach to learning binary codes for large-scale image retrieval. IEEE transactions on pattern analysis and machine intelligence, 35(12): 2916--2929

    work page 2012

  5. [5]

    Gong, Y.; Pawlowski, M.; Yang, F.; Brandy, L.; Bourdev, L.; and Fergus, R. 2015. Web scale photo hash clustering on a single machine. In Proceedings of the IEEE conference on computer vision and pattern recognition, 19--27

    work page 2015

  6. [6]

    Han, Z.; Zhang, C.; Fu, H.; and Zhou, J. T. 2022. Trusted multi-view classification with dynamic evidential fusion. IEEE transactions on pattern analysis and machine intelligence, 45(2): 2551--2566

    2022

  7. [7]

    Hu, K.; Li, Y.; Zhang, S.; Wu, J.; Gong, S.; Jiang, S.; and Weng, L. 2024 a . FedMMD : A Federated weighting algorithm considering Non - IID and Local Model Deviation . Expert Systems with Applications, 237: 121463

    work page 2024

  8. [8]

    Hu, K.; Shen, C.; Wang, T.; Shen, S.; Cai, C.; Huang, H.; and Xia, M. 2023. Action Recognition Based on Multi - Level Topological Channel Attention of Human Skeleton . Sensors, 23(24): 9738

    work page 2023

Show all 61 references
  1. [9]

    Hu, K.; Zhang, E.; Xia, M.; Wang, H.; Ye, X.; and Lin, H. 2024 b . Cross-dimensional feature attention aggregation network for cloud and snow recognition of high satellite images. Neural Computing and Applications, 36(14): 7779--7798

    2024

  2. [10]

    Jiang, Q.-Y.; and Li, W.-J. 2015. Scalable graph hashing with feature transformation. In International Joint Conference on Artificial Intelligence, volume 15, 2248--2254

    2015

  3. [11]

    Jin, Z.; Li, C.; Lin, Y.; and Cai, D. 2013. Density sensitive hashing. IEEE transactions on cybernetics, 44(8): 1362--1371

    2013

  4. [12]

    Kang, Z.; Guo, Z.; Huang, S.; Wang, S.; Chen, W.; Su, Y.; and Xu, Z. 2019. Multiple partitions aligned clustering. arXiv preprint arXiv:1909.06008

    2019 arXiv

  5. [13]

    Kang, Z.; Zhou, W.; Zhao, Z.; Shao, J.; Han, M.; and Xu, Z. 2020. Large- Scale Multi - View Subspace Clustering in Linear Time . Proceedings of the AAAI Conference on Artificial Intelligence, 34(04): 4412--4419

    2020

  6. [14]

    Kumar, A.; Rai, P.; and Daume, H. 2011. Co-regularized multi-view spectral clustering. Advances in neural information processing systems, 24

    2011

  7. [15]

    Li, M.; Wang, S.; Liu, X.; and Liu, S. 2024 a . Parameter- Free and Scalable Incomplete Multiview Clustering With Prototype Graph . IEEE Transactions on Neural Networks and Learning Systems, 35(1): 300--310

    2024

  8. [16]

    P.; Sun, Y.; Sun, Q.; Sun, Y.; W

    Li, X.; Pan, Y. P.; Sun, Y.; Sun, Q.; Sun, Y.; W. Tsang, I.; and Ren, Z. 2024 b . Incomplete Multi-view Clustering with Paired and Balanced Dynamic Anchor Learning. IEEE Transactions on Multimedia

    2024

  9. [17]

    P.; Sun, Y.; Sun, Q

    Li, X.; Pan, Y. P.; Sun, Y.; Sun, Q. S.; Tsang, I. W.; and Ren, Z. 2024 c . Fast Unpaired Multi-view Clustering

    2024

  10. [18]

    Li, X.; Ren, Z.; Sun, Q.; and Xu, Z. 2023. Auto-weighted tensor schatten p-norm for robust multi-view graph clustering. Pattern Recognition, 134: 109083

    2023

  11. [19]

    Liu, C.; Wen, J.; Liu, Y.; Huang, C.; Wu, Z.; Luo, X.; and Xu, Y. 2024 a . Masked two-channel decoupling framework for incomplete multi-view weak multi-label learning. Advances in Neural Information Processing Systems, 36: 32387--32400

    2024

  12. [20]

    Liu, J.; Wang, C.; Gao, J.; and Han, J. 2013. Multi-view clustering via joint nonnegative matrix factorization. In Proceedings of the 2013 SIAM international conference on data mining, 252--260

    2013

  13. [21]

    Liu, S.; Liu, X.; Wang, S.; Niu, X.; and Zhu, E. 2024 b . Fast incomplete multi-view clustering with view-independent anchors. IEEE Transactions on Neural Networks and Learning Systems, 1--12

    2024

  14. [22]

    Long, Z.; Zhu, C.; Chen, J.; Li, Z.; Ren, Y.; and Liu, Y. 2023. Multi-view MERA subspace clustering. IEEE Transactions on Multimedia

    2023

  15. [23]

    Lu, C.; Feng, J.; Chen, Y.; Liu, W.; Lin, Z.; and Yan, S. 2019. Tensor robust principal component analysis with a new tensor nuclear norm. IEEE transactions on pattern analysis and machine intelligence, 42(4): 925--938

    2019

  16. [24]

    Lu, Z.; Liu, G.; and Wang, S. 2020. Sparse neighbor constrained co-clustering via category consistency learning. Knowledge-Based Systems, 201: 105987

    2020

  17. [25]

    Lu, Z.; Nie, F.; Wang, R.; and Li, X. 2023 a . A differentiable perspective for multi-view spectral clustering with flexible extension. IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(6): 7087--7098

    2023

  18. [26]

    Lu, Z.; Wang, S.; Liu, G.; and Nie, F. 2023 b . Robust weighted co-clustering with global and local discrimination. Pattern Recognition, 138: 109405

    2023

  19. [27]

    Ng, A.; Jordan, M.; and Weiss, Y. 2001. On spectral clustering: Analysis and an algorithm. Advances in neural information processing systems, 14

    2001

  20. [28]

    Nie, F.; Cai, G.; and Li, X. 2017. Multi-view clustering and semi-supervised classification with adaptive neighbours. In Proceedings of the AAAI conference on artificial intelligence, volume 31

    2017

  21. [29]

    Nie, F.; Li, J.; Li, X.; et al. 2016. Parameter-free auto-weighted multiple graph learning: a framework for multiview clustering and semi-supervised classification. In International Joint Conference on Artificial Intelligence, volume 9, 1881--1887

    2016

  22. [30]

    Nie, F.; Liu, C.; Wang, R.; Wang, Z.; and Li, X. 2021. Fast fuzzy clustering based on anchor graph. IEEE Transactions on Fuzzy Systems, 30(7): 2375--2387

    2021

  23. [31]

    Qin, Y.; Pu, N.; and Wu, H. 2024. EDMC: Efficient Multi-View Clustering via Cluster and Instance Space Learning. IEEE Transactions on Multimedia, 26: 5273--5283

    2024

  24. [32]

    Qin, Y.; Qin, C.; Zhang, X.; and Feng, G. 2024. Dual Consensus Anchor Learning for Fast Multi-View Clustering. IEEE Transactions on Image Processing, 33: 5298--5311

    2024

  25. [33]

    Shen, X.; Liu, W.; Tsang, I.; Shen, F.; and Sun, Q.-S. 2017. Compressed k-means for large-scale clustering. In Proceedings of the AAAI conference on artificial intelligence, volume 31

    2017

  26. [34]

    Sun, Y.; Qin, Y.; Peng, D.; Peng, X.; and Hu, P. 2024. Robust Multi-View Clustering with Noisy Correspondence. IEEE Transactions on Knowledge and Data Engineering, 1--14

    2024

  27. [35]

    Sun, Y.; Ren, Z.; Hu, P.; Peng, D.; and Wang, X. 2023 a . Hierarchical consensus hashing for cross-modal retrieval. IEEE Transactions on Multimedia

    2023

  28. [36]

    Sun, Y.; Wang, X.; Peng, D.; Ren, Z.; and Shen, X. 2023 b . Hierarchical hashing learning for image set classification. IEEE Transactions on Image Processing, 32: 1732--1744

    2023

  29. [37]

    Tian, Z.; Zhang, H.; Chen, Y.; and Zhang, D. 2020. Unsupervised hashing based on the recovery of subspace structures. Pattern Recognition, 103: 107261

    2020

  30. [38]

    Wang, D.; Wang, Q.; and Gao, X. 2017. Robust and flexible discrete hashing for cross-modal similarity search. IEEE transactions on circuits and systems for video technology, 28(10): 2703--2715

    2017

  31. [39]

    Wang, H.; Nie, F.; Huang, H.; and Makedon, F. 2011. Fast nonnegative matrix tri-factorization for large-scale data co-clustering. In Twenty-Second International Joint Conference on Artificial Intelligence. Citeseer

    2011

  32. [40]

    Wang, H.; Yang, Y.; and Liu, B. 2019. GMC: Graph-based multi-view clustering. IEEE Transactions on Knowledge and Data Engineering, 32(6): 1116--1129

    2019

  33. [41]

    Wang, H.; Yao, M.; Jiang, G.; Mi, Z.; and Fu, X. 2023. Graph-collaborated auto-encoder hashing for multiview binary clustering. IEEE Transactions on Neural Networks and Learning Systems

    2023

  34. [42]

    Wang, S.; Liu, X.; Liu, L.; Tu, W.; Zhu, X.; Liu, J.; Zhou, S.; and Zhu, E. 2022 a . Highly-efficient incomplete large-scale multi-view clustering with consensus bipartite graph. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 9776--9785

    2022

  35. [43]

    Wang, Z.; Lu, H.; Jin, J.; and Hu, K. 2022 b . Human Action Recognition Based on Improved Two - Stream Convolution Network . Applied Sciences, 12(12): 5784

    2022

  36. [44]

    Weiss, Y.; Torralba, A.; and Fergus, R. 2008. Spectral hashing. Advances in neural information processing systems, 21

    2008

  37. [45]

    Wen, J.; Liu, C.; Deng, S.; Liu, Y.; Fei, L.; Yan, K.; and Xu, Y. 2023 a . Deep double incomplete multi-view multi-label learning with incomplete labels and missing views. IEEE transactions on neural networks and learning systems

    2023

  38. [46]

    Wen, J.; Xu, G.; Tang, Z.; Wang, W.; Fei, L.; and Xu, Y. 2023 b . Graph regularized and feature aware matrix factorization for robust incomplete multi-view clustering. IEEE Transactions on Circuits and Systems for Video Technology

    2023

  39. [47]

    Wen, J.; Zhang, Z.; Fei, L.; Zhang, B.; Xu, Y.; Zhang, Z.; and Li, J. 2023 c . A Survey on Incomplete Multiview Clustering . IEEE Transactions on Systems, Man, and Cybernetics: Systems, 53(2): 1136--1149

    2023

  40. [48]

    Wu, Z.; Lin, X.; Lin, Z.; Chen, Z.; Bai, Y.; and Wang, S. 2023. Interpretable graph convolutional network for multi-view semi-supervised learning. IEEE Transactions on Multimedia

    2023

  41. [49]

    Xia, W.; Gao, Q.; Wang, Q.; Gao, X.; Ding, C.; and Tao, D. 2023. Tensorized bipartite graph learning for multi-view clustering. IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(4): 5187--5202

    2023

  42. [50]

    Xia, Y.; He, K.; Kohli, P.; and Sun, J. 2015. Sparse projections for high-dimensional binary codes. In Proceedings of the IEEE conference on computer vision and pattern recognition, 3332--3339

    2015

  43. [51]

    Yang, B.; Wang, S.; Markham, A.; and Trigoni, N. 2020. Robust Attentional Aggregation of Deep Feature Sets for Multi -view 3D Reconstruction . International Journal of Computer Vision, 128(1): 53--73

    2020

  44. [52]

    Zhang, H.; Nie, F.; and Li, X. 2023. Large- Scale Clustering With Structured Optimal Bipartite Graph . IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(8): 9950--9963

    2023

  45. [53]

    Zhang, R.; and Lu, Z. 2016. Large Scale Sparse Clustering. In International Joint Conference on Artificial Intelligence, 2336--2342

    2016

  46. [54]

    Zhang, X.; Yang, Y.; Li, T.; Zhang, Y.; Wang, H.; and Fujita, H. 2021. CMC : A consensus multi-view clustering model for predicting Alzheimer ’s disease progression. Computer Methods and Programs in Biomedicine, 199: 105895

    2021

  47. [55]

    Zhang, Y.; Xie, Y.; Li, C.; Wu, Z.; and Qu, Y. 2022. Learning all-in collaborative multiview binary representation for clustering. IEEE Transactions on Neural Networks and Learning Systems

    2022

  48. [56]

    Zhang, Y.; Xie, Y.; Li, C.; Wu, Z.; and Qu, Y. 2024. Learning All - In Collaborative Multiview Binary Representation for Clustering . IEEE Transactions on Neural Networks and Learning Systems, 35(3): 4260--4273

    2024

  49. [57]

    Zhang, Z.; Liu, L.; Qin, J.; Zhu, F.; Shen, F.; Xu, Y.; Shao, L.; and Shen, H. T. 2018 a . Highly-economized multi-view binary compression for scalable image clustering. In Proceedings of the European conference on computer vision (ECCV), 717--732

    2018

  50. [58]

    T.; and Shao, L

    Zhang, Z.; Liu, L.; Shen, F.; Shen, H. T.; and Shao, L. 2018 b . Binary multi-view clustering. IEEE transactions on pattern analysis and machine intelligence, 41(7): 1774--1782

    2018

  51. [59]

    Zhao, J.; Kang, F.; Zou, Q.; and Wang, X. 2023. Multi-view clustering with orthogonal mapping and binary graph. Expert Systems with Applications, 213: 118911

    2023

  52. [60]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all...

  53. [61]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.