- Rating
- 4(+1)
- Confidence
- 4
- Informativeness
- 3
- Presentation
- 3
- Soundness
- 3
Read review
Paper summary
This paper proposes a Low-Rank Smooth Unified Anchor Tensor Factorization method (LS-UATNOF) to address the reliance on predefined anchors and the lack of cross-view anchor consistency in existing methods. The proposed method projects heterogeneous multi-view data into a common feature space to learn a unified anchor matrix, and incorporates low-rank smoothness regularization to capture structural priors. Extensive comparative and ablation experiments demonstrate the effectiveness and competitiveness of LS-UATNOF.
Critical strengths
1 The method is well motivated. It jointly optimizes unified anchors and tensor factorization to enhance cross-view consistency, while integrating low-rank and smoothness priors through tensor total variation Schatten-p regularization, reducing the need for additional hyperparameter tuning. 2 The experimental results are convincing. LS-UATNOF outperforms state-of-the-art methods on multiple benchmark datasets, demonstrating strong clustering performance.
Critical weaknesses
1 The rationale behind the model design is insufficiently discussed. The authors should clarify why Schatten-p regularization is preferred over the standard tensor nuclear norm and explain the advantages of shared anchors over independently learned anchors. 2 The rationale for setting the common latent dimension to (d=min{c,min d_v}), as well as its impact on clustering performance, requires further clarification.
Questions for authors
1 Why do the authors apply regularization to the anchor representation tensor rather than the cluster-indicator tensor? 2 What are the respective roles of the nonnegative and orthogonal tensor factors, and what motivates imposing different constraints on them? 3 Is there a consistent pattern in the optimal values of p across different datasets? Could the authors offer recommendations on how to choose p in practice?
4 comments
Response to Reviewer QvU3 (Part 2/2)
Question 1: Regularizing the anchor representation tensor.
Thank you for this question. We regularize because it directly represents sample-anchor relationships learned from the observed data. Reconstruction alone may retain noise and variations unrelated to clustering; TTV--Schatten- regularization encourages low-rank, smooth variations in these representations and guides the subsequent tensor factorization. By contrast, is the compact clustering representation, whose orthogonality constraint reduces redundancy among its components. These mechanisms serve complementary purposes: improving the structure of the anchor representations and separating components in the final clustering representation.
Question 2: Different constraints on the tensor factors.
We appreciate this question. Nonnegativity on provides a nonnegative basis for expressing anchor-space patterns, whereas orthogonality on reduces redundancy among the components used for clustering. The constraints therefore reflect the different roles of the two factors. Applying them separately also simplifies optimization: is updated through nonnegative projection, while is updated through slice-wise SVD in the Fourier domain, without requiring both constraints to be satisfied by the same factor.
Question 3: Selection of the Schatten parameter .
Thank you for asking about practical parameter selection. The table below summarizes the ACC results underlying Figure 5 in Appendix D, with all other parameters fixed and only varied. The preferred value depends on the dataset, and performance does not improve monotonically with . Nevertheless, matches the best ACC on HW and BDGP and is within 0.05 percentage points of the best on CCV and Animal. We therefore recommend as a practical starting point rather than a universally optimal choice. It recovers the tensor nuclear-norm penalty on the directional gradient tensor while retaining the smoothness prior; smaller values can be explored when additional tuning is affordable.
| HW | BDGP | CCV | Animal | |
|---|---|---|---|---|
| 0.1 | 99.75 | 99.52 | 39.70 | 95.76 |
| 0.2 | 99.85 | 99.52 | 73.79 | 97.20 |
| 0.3 | 99.80 | 99.52 | 87.08 | 90.93 |
| 0.4 | 99.70 | 99.52 | 86.51 | 94.79 |
| 0.5 | 99.70 | 99.52 | 86.31 | 94.78 |
| 0.6 | 99.55 | 99.52 | 79.46 | 88.85 |
| 0.7 | 99.65 | 99.52 | 85.04 | 97.16 |
| 0.8 | 99.90 | 99.52 | 87.17 | 97.15 |
| 0.9 | 100.00 | 99.52 | 87.23 | 97.15 |
| 1.0 | 100.00 | 99.52 | 87.18 | 97.16 |
ACC (%); the best result in each dataset is shown in bold.
15579
The authors have addressed my concerns properly and I prefer to increase the rating score.
Reply to Reviewer QvU3
Thank you for your positive feedback and for increasing your rating. We are pleased that our revisions and responses have addressed your concerns. We will incorporate the relevant revisions and clarifications into the final version of the paper.
Response to Reviewer QvU3 (Part 1/2)
We are grateful to the reviewer for recognizing the motivation of our approach and its clustering performance. We also welcome the questions about the regularization design, latent dimension, and tensor-factor constraints, as well as the request for practical guidance on choosing p. Our point-by-point responses below explain these design choices and discuss the relevant experimental findings.
Weakness 1: Rationale for the model design.
We thank the reviewer for encouraging us to clarify our motivation. For 0<p<1, Schatten-p regularization applies the nonconvex mapping σp to each singular value σ. Its slope pσp−1 decreases as σ increases, placing stronger shrinkage on small singular values while reducing the bias on large ones compared with the nuclear norm. This helps suppress weak components while preserving dominant structures; p=1 recovers the nuclear-norm case.
Our motivation for unified anchors goes beyond cross-view consistency: they provide aligned anchor indices for subsequent tensor factorization. When anchors are learned independently, the same row index in different views' anchor representations may refer to different anchors. Stacking these representations can therefore mix anchor mismatch with meaningful cross-view variation. Sharing A establishes a common anchor reference, allowing tensor factorization to capture cross-view relationships while retaining view-specific representation coefficients.
Weakness 2: Choice of the common latent dimension.
We appreciate the reviewer's question about this setting. The number of clusters c provides a task-related reference for the latent dimension, yielding a compact representation whose size scales with the clustering task. Using it as the default also avoids introducing an additional dimension-search procedure. Since row orthogonality of P(v) requires d≤dv for every view, we set d=minc,minvdv. This provides a practical, task-informed choice while ensuring feasibility across views, rather than assuming a universally optimal dimension.
We further examine the effect of d on BDGP and CCV, fixing m=10 and m=20, respectively, and retaining all other settings except the two regularization parameters. For each dimension, we independently search the same 49 combinations of λ1 and λ2 and report the results of the combination with the highest ACC.
ACC on BDGP; the best results are shown in bold.
ACC on CCV; the best result is shown in bold.
On BDGP, ACC ranges from 99.52 to 99.72, a difference of only 0.20 percentage points; the default d=5 thus remains competitive across the tested dimensions. On CCV, ACC increases from 86.27 at d=10 to 87.18 at the default d=20, which achieves the best result among the tested values. Together, these results support the proposed dimension rule as a practical default, providing strong performance on these two datasets without an additional dataset-specific dimension search.