Pith. sign in

REVIEW 4 major objections 5 minor 52 references

3DeepRep: 3D Deep Low-rank Tensor Representation for Hyperspectral Image Inpainting

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 The paper proposes 3DeepRep, a self-supervised tensor model that enforces low-rankness in three learned latent spaces—one per mode—and claims it outperforms six existing methods on hyperspectral image inpainting.

desk verdict 3DeepRep reports large gains for HSI inpainting from a 3-directional deep low-rank representation, but the paper doesn't show the low-rank penalty is what does the work, and it skips the closest prior. read the letter →

arxiv 2506.16735 v1 pith:FPFXZZM7 submitted 2025-06-20 cs.CV eess.IV

Yunshan Li , Wenwu Gong , Qianqian Wang , Chao Wang , Lili Yang This is my paper · ORCID
classification cs.CVeess.IV
keywords hyperspectralimageinpaintingtensornuclearnormdeeptransformlow-rankrepresentationself-supervisedlearning3-directionalcoupledblock
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 3DeepRep, a self-supervised model for filling missing entries in hyperspectral images. It claims that imposing a tensor nuclear norm prior in three learned latent spaces—one per tensor mode—captures spatial and spectral structure that single-direction deep transforms miss. The three reconstructions are fused by a learned aggregation module, and the whole model is trained by gradient descent using only the observed pixels. The authors report that 3DeepRep outperforms six existing methods on three simulated datasets and a real MODIS scene across four missing-data patterns.

What carries the argument

The central object is the 3-directional tensor nuclear norm (3DTNN) regularization: for each mode $i=1,2,3$, a deep transform $g_i$ maps a latent tensor into image space, and low-rankness is promoted by summing the matrix nuclear norms of the mode-$i$ frontal slices of the latent tensor. The transform $g_i$ is built from a coupled transform block—a slice-wise $3\times3$ convolution followed by a fully connected layer along the mode—so each branch captures local spatial correlations and global mode-wise correlations. The aggregation module $G$ applies the same block to the concatenation of the three directional reconstructions. The model minimizes the three directional losses plus a final observed-pixel fidelity term, and the paper derives subgradients for the nuclear norm terms to justify Adam-based training. The latent tensors are learned directly, which the paper reports reduces parameters and improves accuracy relative to generator-based latent representations.

What would settle it

Train 3DeepRep on a partially observed HSI where the missing region contains fine structure that is not low-rank along any mode (for example, random high-frequency texture), and compare against a model that only fits observed pixels. If the missing-region error is no lower, or if replacing the TNN initialization of the latent tensors with random tensors preserves the reported margins over CoNoT, then the 3-directional low-rank representation is not the operative mechanism.

Watch

Extended reading notes

Core claim

3DeepRep's central claim is that low-rankness should be enforced not only along the spectral mode but along all three modes of the HSI tensor, each through its own deep nonlinear transform. For each direction $i=1,2,3$, the model permutes the tensor so mode $i$ becomes the third mode, maps it to a latent tensor with a coupled transform block (a 2D convolution on frontal slices followed by a fully connected layer along the mode), and penalizes the sum of nuclear norms of the latent tensor's frontal slices. The three directional reconstructions are then aggregated by another coupled transform block into the final image. The paper argues that this 3-directional TNN regularization yields better inpainting than spectral-only deep TNN methods, and reports consistent gains over TNN, WSTNN, t-CTV, S2NTNN, CoNoT, and HIR-Diff on Pavia Centre, Pavia University, Washington DC Mall, and a MODIS band.

Load-bearing premise

The load-bearing premise is that low-rankness in three simultaneously learned latent spaces is a genuine image prior that transfers to missing regions; because the same observed-pixel fidelity used to train the output also trains the latent tensors, the optimizer could in principle fit the observed pixels without learning any valid tensor prior, leaving generalization to rest on the subgradient training and the TNN-based initialization.

Editorial extensions

If this is right

  • If the central claim is correct, HSI inpainting no longer needs to commit the low-rank prior to a single mode; the same tensor can be low-rank in three different learned coordinate systems at once.
  • The learned aggregation replaces the hand-tuned balancing hyperparameters of ADMM-based three-directional TNN methods, lowering computational cost and tuning risk.
  • Learning the latent tensors directly, rather than generating them from a network, is both cheaper and more accurate, so future deep low-rank models can drop the generator branch.
  • The method's self-supervised formulation means it applies to any partially observed HSI without ground truth, including real sensor scenes such as the MODIS band shown in the paper.
  • The reported margin in the wide-deadline case (PSNR 51.052 versus 43.499 for the second-best method on Pavia Centre) is the strongest quantitative evidence that multi-directional structure, not just spectral redundancy, drives the improvement.

Reading between the lines

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

  • The dependence on TNN initialization suggests the deep branches refine a classical low-rank solution rather than discovering structure from scratch; a randomized-initialization ablation would test this directly.
  • Because the mode-3 (spectral) branch alone beats either spatial branch alone, the value of adding directions may saturate; comparing a 6-direction variant (including diagonal or patch modes) would show whether directional coverage has diminishing returns.
  • The framework transfers naturally to other three-mode inverse problems, such as video or multi-spectral denoising, where the same permutation-plus-deep-transform template applies.
  • The reported insensitivity to gamma (intermediate fidelity weight) hints that the nuclear-norm regularizer, not the per-branch fidelity, carries the prior; setting gamma=0 and checking whether the final fidelity term alone retains performance would isolate that contribution.
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 3DeepRep, a self-supervised model for hyperspectral image (HSI) inpainting that extends transform-based tensor nuclear norm (TNN) methods to three tensor modes. The model maintains three learnable latent tensors \hat{X}_i, one per direction, enforces low-rankness by summing the nuclear norms of mode-i frontal slices of each latent tensor, maps each latent tensor back to the image domain through a coupled transform block (CNN plus FC along the corresponding mode), and fuses the three directional reconstructions with a learned aggregation module. The optimization is performed with Adam using a subgradient of the nuclear norm terms. Experiments on three simulated HSI datasets (Pavia Centre, Pavia University, Washington DC Mall) and one real MODIS scene compare against TNN, WSTNN, t-CTV, S2NTNN, CoNoT, and HIR-Diff across four missing-data scenarios, reporting consistently higher PSNR/SSIM and lower SAM in most settings. Ablations examine the number of directions, learnable versus generated latent tensors, expansion ratio, and hyperparameter sensitivity.

Significance. If the empirical claims hold, the paper demonstrates a useful extension of TNN-based deep transforms: applying directional deep transforms along all three modes with a learned aggregation is plausible to capture spatial and spectral low-rank structure more fully than single-mode transforms. The work has clear strengths: the method is self-supervised and does not require ground truth; the derivation of the nuclear-norm subgradient is standard and the algorithm is concretely specified; the ablation study comparing 1D versus 3D directions and learnable versus generated latent tensors directly targets design choices; and the evaluation includes a real-data MODIS scene in addition to simulated experiments. However, the evaluation protocol contains a serious selection-bias issue (hyperparameters tuned to maximize PSNR on the same test images), omits the closest prior three-directional nonlinear TNN method [38] from comparisons, and provides no variability estimates for random missing masks. Because the reported margins are the main evidence for superiority, these issues are load-bearing rather than cosmetic.

major comments (4)
  1. [Section V-A3] The hyperparameters β, θ, and γ are selected by maximizing PSNR on the same test images that are later reported in Tables II–IV, with no separate validation set. This introduces selection bias: the reported numbers are in-sample optima rather than unbiased estimates of generalization. The comparison baselines, in contrast, use hyperparameters 'manually adjusted following the authors' suggestions.' Please use a validation-based selection protocol (e.g., a held-out portion of the image or a separate tuning set), or report performance across a hyperparameter grid for all methods, so that the margins in Tables II–IV can be interpreted fairly.
  2. [Section IV-C, Remark; Section V] Reference [38] (self-supervised nonlinear transform based three-directional tensor nuclear norm) is explicitly discussed in the Remark as a prior 3DTNN-based method, and the paper's stated novelty is the 3-directional deep transform framework. Yet [38] is not included in any experimental comparison, while more distant methods such as HIR-Diff are. This omission is particularly concerning because the claim of 'superior performance compared to existing state-of-the-art approaches' should be evaluated against the method that is closest to the proposed model. Please include [38] in the comparisons, or if its code is unavailable, state this and justify why the comparison is still representative.
  3. [Section V-C3, Table V, Eq. (20)] The paper does not establish that the 3-directional low-rank penalty is the active ingredient in the reported gains. The latent tensors are free variables of full image dimension (about 9.6M of the 9.712M parameters in the 3D model, per Table V), the reported β values are as small as 10^-6, and for Washington DC Mall the optimal γ is zero. There is no ablation with β=0 (or with the nuclear-norm terms removed) and no report of the rank or nuclear norm of the learned \hat{X}_i relative to their TNN initialization. Consequently, the improvements could plausibly be driven by the deep transforms' implicit bias and the TNN initialization rather than by the low-rank regularizer. Please add a β=0 control, report the nuclear norms and approximate ranks of the optimized latent tensors, and adjust the conceptual claims if the low-rank term is not load-bearing.
  4. [Section V-A2, Tables II–IV] Case 1 uses 'randomly select elements' as the missing mask, and Case 4 combines random point, stripe, and deadline masks, but the reported tables contain single numbers with no indication of the random seed, the number of random draws, or variance across draws. For random masks, a single draw makes it impossible to know whether the reported PSNR differences are statistically meaningful. Please report mean and standard deviation over at least a small number of random mask realizations (or random seeds for the deep models) for the stochastic cases.
minor comments (5)
  1. [Section V-A3] The text says the hyperparameters to tune are β, θ, and λ, but the objective (20) is written in terms of β, γ, and θ with λ_i = γ/α_i; please keep the notation consistent throughout the parameter-setting discussion.
  2. [Section V-B] For the real MODIS experiment, hyperparameters of all methods were adjusted 'to achieve the best visual result,' which is subjective and not reproducible; please provide a fixed protocol or quantitative selection criterion.
  3. [Definition 2, Section III] The concatenation formula contains garbled index expressions, e.g., 's=s3+n1 3+n2 3', which appear intended to be s = s3, s3+n3, s3+2n3; please correct the typographical corruption.
  4. [Eq. (11) and surrounding text] The symbols l, j, s are used both as free indices of a particular entry of X_i and as summation indices in the same expression; please rename the summation indices to avoid ambiguity.
  5. [Fig. 8] The sensitivity-analysis curves are described only by the caption 'Sensitivity analysis of hyperparameters'; please add axis labels, a legend identifying each dataset, and the ranges of β, θ, and γ tested.

Circularity Check

1 steps flagged · score 2.0 of 10

Mild test-set tuning circularity; central 3DeepRep derivation is self-contained.

  1. fitted input called prediction [Section V-A3 (Parameter settings), Eq. (20); Section V-C3 (Hyperparameter Analysis), Fig. 8]
    "We adjust hyperparameters β, θ, λ to achieve the highest PSNR value, with β controlling the importance of the low-rankness regularizers, θ balancing the importance of low-rankness between the spatial mode (mode-1 and mode-2) and the spectral mode (mode-3), γ determining the importance of fidelity of Xi’s, as illustrated in (20):"

    The evaluation protocol selects hyperparameters by maximizing PSNR on the same images whose PSNR is then reported as the method's performance, and the sensitivity analysis states per-dataset optima (e.g., β=10−5, 10−5, 10−6 for the three datasets). The reported scores therefore are the result of fitting the evaluation configuration to the test data, not predictions of a prespecified method on unseen data; in that narrow sense the headline quantitative claim is enforced by the tuning. The derivation in Eqs. (8)–(10) itself is not circular, and no fitted parameter is renamed as a new theoretical quantity.

full rationale

The central model equations (8)–(10) are not circular: the objective is a standard self-supervised regularized inversion combining nuclear-norm penalties on three learnable latent tensors with observed-pixel fidelity on the directional reconstructions and the aggregated output, and the reported inpainting quality is measured on missing entries that do not appear in the fidelity terms. The deep transform and aggregation use the published CTB module from CoNoT [37] by citation, but that citation supplies an architectural building block rather than the paper's novelty, which is the three-directional extension; no load-bearing uniqueness theorem or self-citation chain forces the result. The initialization of the latent tensors by conventional TNN is an initialization choice, not a fitted prediction. The only defensible circularity concern is the evaluation protocol in Section V-A3, where hyperparameters are adjusted to maximize PSNR on the test images and the sensitivity analysis reports per-dataset optima, so the headline comparison numbers are partially selected on the evaluation target. This is a real but mild evaluation circularity; it does not make the model derivation equivalent to its inputs.

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

The central claim rests on the assumption that hyperspectral images are low-rank in three learned latent spaces, on the expressiveness of the coupled transform block as an inverse mapping, on the subgradient characterization of the nuclear norm, and on the aggregation network's ability to fuse directional reconstructions. The fitted quantities in the model are the weighting hyperparameters beta, theta, and gamma, plus the expansion ratio k; these are selected per dataset to maximize PSNR, which weakens the reported performance as a prediction.

free parameters (4)
  • beta (direction weight scale) = 1e-5 (Pavia Centre, Pavia University), 1e-6 (Washington DC Mall)
    Weights the three directional low-rank regularizers in the final loss; tuned per dataset to maximize PSNR (Section V-C3).
  • theta (spatial vs spectral low-rank balance) = 0.1 (Pavia Centre, Pavia University), 1.0 (Washington DC Mall)
    Sets the directional weights as proportional to (1,1,theta) and is tuned per dataset to maximize PSNR.
  • gamma (directional fidelity weight) = 1e-7 (Pavia Centre), 1e-8 (Pavia University), 0 (Washington DC Mall)
    Controls the fidelity of the three directional reconstructions; tuned per dataset, with an optimal value of zero for Washington DC Mall.
  • k (latent expansion ratio) = 1 (chosen for main experiments)
    Latent dimensions are set to k times the image dimensions; k=1 is chosen to reduce computational cost even though k=2 gives a slightly higher PSNR.
assumptions (4)
  • standard math The subgradient of the nuclear norm can be computed by choosing W=0 in the subdifferential characterization of Theorem 3.
    Used in Section IV-D to derive the gradient descent updates for the latent tensors.
  • domain assumption Hyperspectral image tensors admit low-rank structure in latent spaces along all three tensor modes.
    Central modeling premise in Sections I and IV; if false, the three directional nuclear norm regularizers do not capture useful structure.
  • domain assumption The coupled transform block, a slice-wise 2D convolution plus a fully connected layer, is expressive enough to serve as an inverse mapping from each low-rank latent tensor to the image domain.
    Defines g_i in Eq. (6); no invertibility or approximation guarantee is provided.
  • ad hoc to paper The aggregation function G = g3(concat(...)) can fuse three directional reconstructions without introducing artifacts or additional hand-tuned balancing weights.
    Introduced in Eq. (7); its validity is demonstrated only empirically through the final reconstruction quality.

how reviews work

0 comments
Cite this review

Pith. "Pith review of 3DeepRep: 3D Deep Low-rank Tensor Representation for Hyperspectral Image Inpainting." pith.science (2026). https://pith.science/paper/FPFXZZM7

@misc{pith2026250616735,
  author       = {Pith},
  title        = {Pith review of: 3DeepRep: 3D Deep Low-rank Tensor Representation for Hyperspectral Image Inpainting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FPFXZZM7}},
  note         = {Machine review of arXiv:2506.16735}
}
read the original abstract

Recent approaches based on transform-based tensor nuclear norm (TNN) have demonstrated notable effectiveness in hyperspectral image (HSI) inpainting by leveraging low-rank structures in latent representations. Recent developments incorporate deep transforms to improve low-rank tensor representation; however, existing approaches typically restrict the transform to the spectral mode, neglecting low-rank properties along other tensor modes. In this paper, we propose a novel 3-directional deep low-rank tensor representation (3DeepRep) model, which performs deep nonlinear transforms along all three modes of the HSI tensor. To enforce low-rankness, the model minimizes the nuclear norms of mode-i frontal slices in the corresponding latent space for each direction (i=1,2,3), forming a 3-directional TNN regularization. The outputs from the three directional branches are subsequently fused via a learnable aggregation module to produce the final result. An efficient gradient-based optimization algorithm is developed to solve the model in a self-supervised manner. Extensive experiments on real-world HSI datasets demonstrate that the proposed method achieves superior inpainting performance compared to existing state-of-the-art techniques, both qualitatively and quantitatively.

Figures

Figures reproduced from arXiv: 2506.16735 by the authors.

Figure 1
Figure 1. Qualitative and quantitative comparison of inpainting models: 1D-D [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Flowchart of the proposed 3DeepRep method for HSI inpainting. The hyperspectral image is encoded into three low-rank latent tensors, where [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The four deadline patterns on the Pavia Centre dataset. (a) The pattern [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Inpainting results of the Pavia Centre dataset under Case 4 (mixture missing). The top row is the false-color images composed of bands (R: 70, G: [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Inpainting results of the Pavia University dataset under Case 4 (mixture missing). The top row is the false-color images composed of bands (R: 70, [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: Inpainting results of the Washington DC Mall dataset under Case 4 (mixture missing). The top row is the false-color images composed of bands (R: [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: The inpainting results on Band 27 of the chosen MODIS Level-1B HSI dataset which is with the 1 km resolution at Antarctic Peninsula on February [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 8
Figure 8. Figure 8: Sensitivity analysis of hyperparameters. [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

52 extracted references · 47 canonical work pages

  1. [38]

    Multi-Dimensional Image Recovery Via Self-Supervised Nonlinear Transform Based a Three- Directional Tensor Nuclear Norm,

    G. Li, Z. Tu, J. Lu, C. Wang, and L. Shen, “Multi-Dimensional Image Recovery Via Self-Supervised Nonlinear Transform Based a Three- Directional Tensor Nuclear Norm,” Numerical Mathematics: Theory, Methods and Applications , 2024

    work page 2024

  2. [1]

    Modern trends in hyperspectral image analysis: A review,

    M. J. Khan, H. S. Khan, A. Yousaf, K. Khurshid, and A. Abbas, “Modern trends in hyperspectral image analysis: A review,” IEEE Access, vol. 6, pp. 14118–14129, 2018

    work page 2018

  3. [2]

    Accurate hyperspectral imaging of mineralised out- crops: An example from lithium-bearing pegmatites at uis, namibia,

    R. Booysen, S. Lorenz, S. T. Thiele, W. Fuchsloch, T. Marais, P. Nex, and R. Gloaguen, “Accurate hyperspectral imaging of mineralised out- crops: An example from lithium-bearing pegmatites at uis, namibia,” Remote Sensing of Environment , 2021

    work page 2021

  4. [3]

    A systematic review of hyperspectral imaging in precision agriculture: Analysis of its current state and future prospects,

    B. G. Ram, P. Oduor, C. Igathinathane, K. Howatt, and X. Sun, “A systematic review of hyperspectral imaging in precision agriculture: Analysis of its current state and future prospects,” Comput. Electron. Agric., vol. 222, p. 109037, 2024

    work page 2024

  5. [4]

    Target detection through tree-structured encoding for hyperspectral images,

    X. Sun, Y . Qu, L. Gao, X. Sun, H. Qi, B. Zhang, and T. Shen, “Target detection through tree-structured encoding for hyperspectral images,” IEEE Transactions on Geoscience and Remote Sensing, vol. 59, pp. 4233–4249, 2021

    work page 2021

  6. [5]

    An overview on spectral and spatial information fusion for hyperspectral image classification: Current trends and challenges,

    M. Imani and H. Ghassemian, “An overview on spectral and spatial information fusion for hyperspectral image classification: Current trends and challenges,” Inf. Fusion, vol. 59, pp. 59–83, 2020

    work page 2020

  7. [6]

    Multimodal hyperspectral unmixing: Insights from attention networks,

    Z. Han, D. Hong, L. Gao, J. Yao, B. Zhang, and J. Chanussot, “Multimodal hyperspectral unmixing: Insights from attention networks,” IEEE Transactions on Geoscience and Remote Sensing , vol. 60, pp. 1– 13, 2022

    work page 2022

  8. [7]

    Hyperspectral imaging and its applications: A review,

    A. Bhargava, A. Sachdeva, K. Sharma, M. H. Alsharif, P. Uthansakul, and M. Uthansakul, “Hyperspectral imaging and its applications: A review,” Heliyon, vol. 10, 2024

    work page 2024

Show all 52 references
  1. [8]

    The twist tensor nuclear norm for video completion,

    W. Hu, D. Tao, W. Zhang, Y . Xie, and Y . Yang, “The twist tensor nuclear norm for video completion,” IEEE Transactions on Neural Networks and Learning Systems, vol. 28, pp. 2961–2973, 2017

    2017

  2. [9]

    Tensor completion for estimating missing values in visual data,

    J. Liu, P. Musialski, P. Wonka, and J. Ye, “Tensor completion for estimating missing values in visual data,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 35, pp. 208–220, 2009

    2009

  3. [10]

    Tensor robust principal component analysis with a new tensor nuclear norm,

    C. Lu, J. Feng, Y . Chen, W. Liu, Z. Lin, and S. Yan, “Tensor robust principal component analysis with a new tensor nuclear norm,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 42, pp. 925–938, 2018

    2018

  4. [11]

    Tensor decompositions and applications,

    T. G. Kolda and B. W. Bader, “Tensor decompositions and applications,” SIAM Rev., vol. 51, pp. 455–500, 2009

    2009

  5. [12]

    Low-rank tensor models for improved multidimensional mri: Application to dynamic cardiac $t 1$ mapping,

    B. Yaman, S. Weing ¨artner, N. Kargas, N. Sidiropoulos, and M. Akc ¸akaya, “Low-rank tensor models for improved multidimensional mri: Application to dynamic cardiac $t 1$ mapping,” IEEE Transactions on Computational Imaging , vol. 6, pp. 194–207, 2020

    2020

  6. [13]

    Fusing hyperspectral and multispectral images via coupled sparse tensor factorization,

    S. Li, R. Dian, L. Fang, and J. M. Bioucas-Dias, “Fusing hyperspectral and multispectral images via coupled sparse tensor factorization,” IEEE Transactions on Image Processing , vol. 27, pp. 4118–4130, 2018

    2018

  7. [14]

    Self- supervised nonlinear transform-based tensor nuclear norm for multi- dimensional image recovery,

    Y .-S. Luo, X.-L. Zhao, T.-X. Jiang, Y . Chang, M. K. Ng, and C. Li, “Self- supervised nonlinear transform-based tensor nuclear norm for multi- dimensional image recovery,” IEEE Transactions on Image Processing , vol. 31, pp. 3793–3808, 2022

    2022

  8. [15]

    Hyperspectral image super-resolution via subspace- based low tensor multi-rank regularization,

    R. Dian and S. Li, “Hyperspectral image super-resolution via subspace- based low tensor multi-rank regularization,” IEEE Transactions on Image Processing, vol. 28, pp. 5135–5146, 2019

    2019

  9. [16]

    Analysis of individual differences in multidimensional scaling via an n-way generalization of “eckart-young

    J. D. Carroll and J. J. Chang, “Analysis of individual differences in multidimensional scaling via an n-way generalization of “eckart-young” decomposition,” Psychometrika, vol. 35, pp. 283–319, 1970. This work has been submitted to the ieee for possible publication. Copyright ...

    1970

  10. [17]

    Foundations of the parafac procedure: Models and conditions for an

    R. A. Harshman, “Foundations of the parafac procedure: Models and conditions for an ”explanatory” multi-model factor analysis,” 1970

    1970

  11. [18]

    Some mathematical notes on three-mode factor analysis,

    L. R. Tucker, “Some mathematical notes on three-mode factor analysis,” Psychometrika, vol. 31, pp. 279–311, 1966

    1966

  12. [19]

    Tensor-train decomposition,

    I. Oseledets, “Tensor-train decomposition,” SIAM J. Sci. Comput. , vol. 33, pp. 2295–2317, 2011

    2011

  13. [20]

    Tensor ring decomposition,

    Q. Zhao, G. Zhou, S. Xie, L. Zhang, and A. Cichocki, “Tensor ring decomposition,” ArXiv, vol. abs/1606.05535, 2016

    2016 arXiv

  14. [21]

    Fully- connected tensor network decomposition and its application to higher- order tensor completion,

    Y .-B. Zheng, T. Huang, X. Zhao, Q. Zhao, and T.-X. Jiang, “Fully- connected tensor network decomposition and its application to higher- order tensor completion,” in AAAI Conference on Artificial Intelligence , 2021

    2021

  15. [22]

    Tensor completion via fully-connected tensor network decomposition with regularized factors,

    Y .-B. Zheng, T. Huang, X. Zhao, and Q. Zhao, “Tensor completion via fully-connected tensor network decomposition with regularized factors,” Journal of Scientific Computing , vol. 92, 2022

    2022

  16. [23]

    Low-tubal-rank tensor completion using alternating minimization,

    X.-Y . Liu, S. Aeron, V . Aggarwal, and X. Wang, “Low-tubal-rank tensor completion using alternating minimization,” IEEE Transactions on Information Theory , vol. 66, pp. 1714–1737, 2016

    2016

  17. [24]

    Low-rank preserving t- linear projection for robust image feature extraction,

    X. Xiao, Y . Chen, Y . jiao Gong, and Y . Zhou, “Low-rank preserving t- linear projection for robust image feature extraction,” IEEE Transactions on Image Processing , vol. 30, pp. 108–120, 2020

    2020

  18. [25]

    Nonlocal patch tensor sparse representation for hyperspectral image super-resolution,

    Y . Xu, Z. Wu, J. Chanussot, and Z. Wei, “Nonlocal patch tensor sparse representation for hyperspectral image super-resolution,” IEEE Transactions on Image Processing , vol. 28, pp. 3034–3047, 2019

    2019

  19. [26]

    Novel methods for multilinear data completion and de-noising based on tensor-svd,

    Z. Zhang, G. Ely, S. Aeron, N. Hao, and M. E. Kilmer, “Novel methods for multilinear data completion and de-noising based on tensor-svd,” 2014 IEEE Conference on Computer Vision and Pattern Recognition , pp. 3842–3849, 2014

    2014

  20. [27]

    Mixed noise removal in hyperspectral image via low-fibered-rank regularization,

    Y .-B. Zheng, T.-Z. Huang, X.-L. Zhao, T.-X. Jiang, T.-H. Ma, and T.-Y . Ji, “Mixed noise removal in hyperspectral image via low-fibered-rank regularization,” IEEE Transactions on Geoscience and Remote Sensing , vol. 58, no. 1, pp. 734–749, 2019

    2019

  21. [28]

    H2tf for hyperspectral image denoising: Where hierarchical nonlinear transform meets hier- archical matrix factorization,

    J. Li, J. Xie, Y . Luo, X. Zhao, and J. Wang, “H2tf for hyperspectral image denoising: Where hierarchical nonlinear transform meets hier- archical matrix factorization,” IEEE Geoscience and Remote Sensing Letters, vol. 20, pp. 1–5, 2023

    2023

  22. [29]

    Low-rank tensor completion with a new tensor nuclear norm induced by invertible linear transforms,

    C. Lu, X. Peng, and Y . Wei, “Low-rank tensor completion with a new tensor nuclear norm induced by invertible linear transforms,” 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 5989–5997, 2019

    2019

  23. [30]

    A fast algorithm for co- sine transform based tensor singular value decomposition,

    W.-H. Xu, X. Zhao, and M. K. P. Ng, “A fast algorithm for co- sine transform based tensor singular value decomposition,” ArXiv, vol. abs/1902.03070, 2019

    1902 arXiv

  24. [31]

    Dct based weighted adaptive multi-linear data completion and denoising,

    B. Madathil and S. N. George, “Dct based weighted adaptive multi-linear data completion and denoising,” Neurocomputing, vol. 318, pp. 120–136, 2018

    2018

  25. [32]

    Robust tensor completion using transformed tensor singular value decomposition,

    G.-J. Song, M. K. P. Ng, and X. Zhang, “Robust tensor completion using transformed tensor singular value decomposition,” Numerical Linear Algebra with Applications , vol. 27, 2020

    2020

  26. [33]

    Framelet represen- tation of tensor nuclear norm for third-order tensor completion,

    T.-X. Jiang, M. K. P. Ng, X. Zhao, and T. Huang, “Framelet represen- tation of tensor nuclear norm for third-order tensor completion,” IEEE Transactions on Image Processing , vol. 29, pp. 7233–7244, 2019

    2019

  27. [34]

    Tensor q-rank: new data dependent definition of tensor rank,

    H. Kong, C. Lu, and Z. Lin, “Tensor q-rank: new data dependent definition of tensor rank,” Machine Learning, vol. 110, pp. 1867 – 1900, 2019

    1900

  28. [35]

    Dictionary Learning With Low-Rank Coding Coefficients for Tensor Completion,

    T.-X. Jiang, X. Zhao, H. Zhang, and M. K. P. Ng, “Dictionary Learning With Low-Rank Coding Coefficients for Tensor Completion,” IEEE Transactions on Neural Networks and Learning Systems , vol. 34, pp. 932–946, 2020

    2020

  29. [36]

    Nonlinear Trans- form Induced Tensor Nuclear Norm for Tensor Completion,

    B.-Z. Li, X. Zhao, T.-Y . Ji, X. Zhang, and T. Huang, “Nonlinear Trans- form Induced Tensor Nuclear Norm for Tensor Completion,” Journal of Scientific Computing, vol. 92, 2021

    2021

  30. [37]

    Conot: Coupled nonlinear transform-based low-rank tensor representation for multidimensional image completion,

    J.-L. Wang, T.-Z. Huang, X.-L. Zhao, Y .-S. Luo, and T.-X. Jiang, “Conot: Coupled nonlinear transform-based low-rank tensor representation for multidimensional image completion,” IEEE Transactions on Neural Networks and Learning Systems , 2022

    2022

  31. [39]

    Tensor n-tubal rank and its convex relaxation for low-rank tensor recovery,

    Y .-B. Zheng, T. Huang, X. Zhao, T.-X. Jiang, T.-Y . Ji, and T.-H. Ma, “Tensor n-tubal rank and its convex relaxation for low-rank tensor recovery,” Inf. Sci., vol. 532, pp. 170–189, 2018

    2018

  32. [40]

    Tensor completion network for visual data,

    X. Wang, X.-P. Li, N. D. Sidiropoulos, and H. C. So, “Tensor completion network for visual data,” IEEE Transactions on Signal Processing , vol. 73, pp. 386–400, 2025

    2025

  33. [41]

    U-net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” ArXiv, vol. abs/1505.04597, 2015

    2015 arXiv

  34. [42]

    Compressive spectral image reconstruction using deep prior and low-rank tensor representation.,

    J. Bacca, Y . Fonseca, and H. Arguello, “Compressive spectral image reconstruction using deep prior and low-rank tensor representation.,” Applied optics, vol. 60 14, pp. 4197–4207, 2021

    2021

  35. [43]

    Hlrtf: Hierarchical low- rank tensor factorization for inverse problems in multi-dimensional imaging,

    Y . Luo, X. Zhao, D. Meng, and T.-X. Jiang, “Hlrtf: Hierarchical low- rank tensor factorization for inverse problems in multi-dimensional imaging,” 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 19281–19290, 2022

    2022

  36. [44]

    Deep fully-connected tensor network decomposition for multi-dimensional signal recovery,

    R. Su, X.-L. Zhao, W.-H. Wu, S. Liu, and J. He, “Deep fully-connected tensor network decomposition for multi-dimensional signal recovery,” Signal Process., vol. 233, p. 109903, 2025

    2025

  37. [45]

    Tensor completion via joint reweighted tensor q-nuclear norm for visual data recovery,

    X. Cheng, W. Kong, X. Luo, W. Qin, F. Zhang, and J. Wang, “Tensor completion via joint reweighted tensor q-nuclear norm for visual data recovery,” Signal Process., vol. 219, p. 109407, 2024

    2024

  38. [46]

    Low-rank approximation via generalized reweighted iterative nuclear and frobenius norms,

    Y . Huang, G. Liao, Y . Xiang, L. Zhang, J. Li, and A. Nehorai, “Low-rank approximation via generalized reweighted iterative nuclear and frobenius norms,” IEEE Transactions on Image Processing , vol. 29, pp. 2244– 2257, 2020

    2020

  39. [47]

    Delving deep into rectifiers: Surpassing human-level performance on imagenet classification,

    K. He, X. Zhang, S. Ren, and J. Sun, “Delving deep into rectifiers: Surpassing human-level performance on imagenet classification,” in Proceedings of the IEEE International Conference on Computer Vision , pp. 1026–1034, 2015

    2015

  40. [48]

    Characterization of the subdifferential of some matrix norms,

    G. A. Watson, “Characterization of the subdifferential of some matrix norms,” Linear Algebra and its Applications , vol. 170, pp. 33–45, 1992

    1992

  41. [49]

    Combining weighted total variation and deep image prior for natural and medical image restoration via admm,

    P. Cascarano, A. Sebastiani, M. C. Comes, G. Franchini, and F. Porta, “Combining weighted total variation and deep image prior for natural and medical image restoration via admm,” 2021 21st International Conference on Computational Science and Its Applications (ICCSA) , pp. 39...

    2021

  42. [50]

    Exact tensor completion using t-svd,

    Z. Zhang and S. Aeron, “Exact tensor completion using t-svd,” IEEE Transactions on Signal Processing , vol. 65, pp. 1511–1526, 2015

    2015

  43. [51]

    Guaranteed tensor recovery fused low-rankness and smoothness,

    H. Wang, J. Peng, W. Qin, J. Wang, and D. Meng, “Guaranteed tensor recovery fused low-rankness and smoothness,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 45, pp. 10990–11007, 2023

    2023

  44. [52]

    Hir-diff: Unsupervised hyperspectral image restoration via improved diffusion models,

    L. Pang, X. Rui, L. Cui, H. Wang, D. Meng, and X. Cao, “Hir-diff: Unsupervised hyperspectral image restoration via improved diffusion models,” 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pp. 3005–3014, 2024

    2024

Pith tools

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