Pith. sign in

REVIEW 3 major objections 5 minor 50 references

Tensor Stochastic Regression for High-dimensional Time Series via CP Decomposition

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A CP low-rank coefficient tensor makes tensor time-series regression tractable at sample size proportional to the sum of mode dimensions.

desk verdict A genuinely new CP tensor stochastic regression framework with real theoretical content, but the exact-rank assumption and the theorem's sample complexity are load-bearing, and the paper's own empirical studies operate outside that regime. read the letter →

arxiv 2506.06919 v1 pith:F3IM4VBL submitted 2025-06-07 stat.ME

Shibo Li , Yao Zheng This is my paper
classification stat.ME MSC 62M1062H1260G10
keywords TensorstochasticregressionCPdecompositionTimeseriesautoregressionMixed-frequencydataLow-rankestimationSparse
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 a unified regression framework in which the coefficient tensor linking a tensor-valued response series to a tensor-valued predictor series is assumed to have a CP decomposition of small rank. This turns what is normally a problem with $p_y q_x$ parameters into one with roughly $R$ times the sum of the mode dimensions, making vector, matrix, and tensor autoregressions estimable in high dimensions. The central theoretical claim is a pair of non-asymptotic error bounds: the CP low-rank estimator recovers the true coefficient tensor at rate $O_p(\sqrt{(\sum_d p_d + \sum_d q_d)\log N / T})$, and the sparse version improves this to $O_p(\sqrt{s\log(p_y q_x)/T})$. The same structure also gives readable factor-loading interpretations, including data-driven month-to-quarter aggregation weights in mixed-frequency data and pairwise mode interactions.

What carries the argument

The central object is the CP decomposition $B = \sum_{r=1}^R \omega_r \beta_{r,1} \circ \cdots \circ \beta_{r,N}$, where the unit-norm loading vectors $\beta_{r,d}$ live on the individual modes of the response and predictor tensors and the weights $\omega_r$ set the importance of each rank-one factor. This representation reduces the parameter count from $p_y q_x$ to $R(1 + \sum_d p_d + \sum_d q_d)$ and creates supervised factor processes $f_{r,t} = X_t \times_{d=1}^n \beta_{r,m+d}$ that summarize the predictor tensor. Estimation is carried out by an alternating minimization algorithm adapted from sparse CP tensor regression, with closed-form rank-one updates in the response loadings and ordinary least squares updates in the predictor loadings, optionally followed by hard thresholding for sparsity. The theoretical proof relies on a martingale concentration bound for the dependent design, Hanson-Wright concentration for the linear process $X_t$, restricted strong convexity of the quadratic loss, and covering-number arguments for CP-rank and $\ell_0$-ball parameter sets.

What would settle it

Simulate the model with $B^* = B_{\text{rank }R} + \delta B_{\text{dense}}$, where $\delta B_{\text{dense}}$ is a full-rank perturbation of norm $\delta$, and fit the rank-$R$ estimator for increasing $T$; the claimed rate predicts an error floor proportional to $\delta$, so a clear plateau well above that floor would falsify the exact-rank assumption.

Watch

Extended reading notes

Core claim

The paper claims that the tensor stochastic regression model $Y_t = \langle B, X_t\rangle + E_t$, with $Y_t$ an $m$-th order tensor and $X_t$ an $n$-th order tensor, becomes feasible and interpretable when $B$ is assumed to admit the CP decomposition $B = \sum_{r=1}^R \omega_r \beta_{r,1} \circ \cdots \circ \beta_{r,N}$. Under exact CP rank $R$, uniqueness up to permutation, a stationary linear process for the predictors, and sub-Gaussian innovations, the CP low-rank least-squares estimator satisfies $\|\hat B - B^*\|_F = O_p(\sqrt{(\sum_{d=1}^m p_d + \sum_{d=1}^n q_d)\log N / T})$ and a matching prediction-error bound of $O_p((\sum_{d=1}^m p_d + \sum_{d=1}^n q_d)\log N / T)$. When each loading vector is also entrywise sparse with total support $s = \prod_d s_d$, the sparse CP estimator attains $O_p(\sqrt{s\log(p_y q_x)/T})$ for estimation and $O_p(s\log(p_y q_x)/T)$ for prediction. These rates hold for tensor autoregressions as a special case, with the stationarity condition expressed through the spectral radius of the companion matrix.

Load-bearing premise

The guarantees rest on the true coefficient tensor having an exact CP decomposition of known rank $R$ that is unique up to permutation; if the true tensor is only approximately low-rank or the chosen $R$ is wrong, the stated error rates need not hold.

Editorial extensions

If this is right

  • When the rank $R$ and the mode dimensions are fixed, the sample size needed for consistent estimation scales with the sum of the mode dimensions rather than the full tensor size $p_y q_x$.
  • The framework includes vector, matrix, and tensor autoregressions as special cases, so the paper effectively introduces a CP-based tensor autoregressive model of general lag order with an explicit stationarity condition.
  • The sparse CP estimator has an error rate governed by the total support size $s$ rather than the full dimensions, and it can beat Tucker-based estimators when the true factor loadings are sparse.
  • Mixed-frequency regression becomes a structured form of unrestricted MIDAS: the loading on the high-frequency mode is estimated from data rather than fixed by a prespecified lag polynomial.
  • Pairwise outer products of loading vectors give interpretable maps of response-response, predictor-predictor, and response-predictor interactions, which the paper demonstrates on macroeconomic and air-pollution data.

Reading between the lines

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

  • The theory assumes the true rank $R$ is known and exact; in practice $R$ is selected by cross-validation, so a consistent rank-selection procedure would close the gap between the idealized rates and applied use.
  • When the true tensor is only approximately CP low-rank, the estimator should still recover the best rank-$R$ approximation, but the stated exact-rate guarantees would need a bias term proportional to the approximation error.
  • The same supervised-factor interpretation could be extended to nowcasting with ragged-edge data, since the high-frequency-mode loading indicates which recent observations matter most for the low-frequency target.
  • A direct test of the interpretability claim would be to check whether the learned temporal-aggregation weight $\beta_{r,3}$ matches known publication-lag patterns in macroeconomic data; a mismatch would suggest the model is capturing a different predictive mechanism.
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

3 major / 5 minor

Summary. The paper proposes a unified tensor stochastic regression framework, Y_t = <B, X_t> + E_t, with the coefficient tensor B assumed to admit a CP decomposition of rank R. The framework covers vector, matrix, and tensor responses and predictors, and includes tensor autoregression as a special case. Two estimators are studied: a CP low-rank least-squares estimator and a sparse CP low-rank estimator with entrywise ℓ0 constraints on the CP factors. The main theoretical results are non-asymptotic Frobenius-norm estimation and prediction error bounds (Theorems 1 and 2) and analogous bounds for tensor autoregressions (Corollaries 1 and 2). The paper also proposes an alternating minimization algorithm, reports simulation evidence, and applies the methods to mixed-frequency macroeconomic data and Taiwan air pollution data, emphasizing interpretable mode-pair interaction patterns.

Significance. If the stated bounds are correct, the paper provides a broad and useful framework: the CP structure gives parsimonious parametrization and interpretable interactions between modes, and the sparse variant offers additional dimension reduction. The supplement contains detailed proofs using standard tools — restricted strong convexity, Hanson-Wright inequalities for linear processes, martingale concentration, and covering-number arguments — which I found largely plausible. The simulations and the comparisons against Tucker-based and vectorized benchmarks are informative and support the computational and interpretive claims. The main caveats are that the theorems require oracle knowledge of the rank and sparsity levels, that the proofs apply to global minimizers over nonconvex sets while the algorithm lacks convergence guarantees, and that the empirical applications operate far outside the sample-size regime assumed by the theorems.

major comments (3)
  1. [Section 3.1, Assumption 1, and Supplement Eq. (S5)] The proof of Theorem 1 starts from inequality (S5), loss(B_hat) ≤ loss(B*), which is valid only if the true tensor B* belongs to the feasible set Γ(R) used in the estimator. This requires the practitioner's rank R to be at least the true CP rank. Assumption 1 postulates an exact CP decomposition with R components, but in the methodology the rank is selected by time-series cross-validation (Section 5.1 and Remark S3), and Section 6 explicitly leaves consistent rank selection to future work. If cross-validation selects R smaller than the true rank, B* is infeasible, inequality (S5) fails, and the stated estimation and prediction rates in Theorems 1 and 2 have no support. The same issue applies to the sparsity levels s_d in Assumption 4, which are also selected by cross-validation without a misspecification theory. Please either prove bounds under rank/sparsity misspecification, or state clearly and prominently that the theoretical results are oracle-type and conditional on correct tuning.
  2. [Sections 5.1 and 5.2] The empirical applications do not satisfy the sample-size condition of the theorems. In the macroeconomic application, T=220, R=40, and ∑_d p_d + ∑_d q_d = 179+112+3 = 294, so the theorem's condition T ≳ (κ2/κ1)^2 R(∑ p_d + ∑ q_d) log N requires at least 40×294×log(3) ≈ 12,900 observations even when (κ2/κ1)^2 = 1. In the air pollution application, T=13, R=4, and the six-mode tensor has ∑ p_d + ∑ q_d = 62, requiring at least 4×62×log(6) ≈ 444 observations under the same optimistic normalization. The empirical sections should not be presented as confirming the theoretical rates; at most they are exploratory illustrations of the method. The authors should add an explicit disclaimer that these applications lie outside the proven validity regime.
  3. [Section S2, Algorithm 1, and Section 6] Theorems 1 and 2 concern global minimizers of nonconvex optimization problems over CP-rank and sparsity constraint sets, but the estimator actually computed in simulations and applications is the output of Algorithm 1, an alternating minimization procedure. The paper does not establish that the algorithm converges to a global minimizer, and Section 6 states that convergence analysis is left for future work. Consequently, there is a gap between the object for which rates are proven and the object delivered by the implementation. The authors should either provide conditions under which the algorithm's limit satisfies the theorem (for example, initialization guarantees or a risk bound for local optima), or state that the theorems apply only to the global estimator and that the simulation results verify that global estimator, not the algorithm output.
minor comments (5)
  1. [Section 3.1] There is a typo: "coveriance matrix" should be "covariance matrix".
  2. [Theorem 1] The probability exponent is printed as 2R(∑_{d=1}^N p_d + 1) log{6(N+1)}; this is ambiguous and should be written as 2R ∑_{d=1}^N (p_d + 1) log{6(N+1)}, with p_{m+d} = q_d.
  3. [Remark 4] The phrase "strictly stationarity" should be "strictly stationary", and the reference "L·utkepohl" has a formatting artifact.
  4. [Section 5.2 versus Supplement S4.2] The main text selects R=4 for the sparse estimator, while the supplement's non-sparse CP estimator is later reported as having R=11. The relationship between these two choices should be clarified.
  5. [Assumption 1] The uniqueness part of Assumption 1, based on the Kruskal-rank condition, is not invoked in the proofs of Theorems 1 and 2, which concern the tensor B_hat rather than its factors. If the assumption is retained for interpretability of the loading vectors, the text should say so explicitly.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the CP-regression rates follow from explicit feasibility and concentration assumptions; self-citations enter only as standard generic lemmas.

full rationale

The derivation chain is not circular. The estimation and prediction error bounds in Theorems 1 and 2 (main text Sections 3.1-3.2; proofs in Supplement S5.3) are obtained by a standard M-estimation argument. The pivotal inequality (S5), used also as (S16) for the sparse case, only requires that both the estimator and the true tensor lie in the same feasible CP-rank or sparsity class; Assumptions 1 and 4 explicitly impose this feasibility (exact rank R and sparsity levels s_d). This is a modelling assumption, not a fitted quantity that is later reported as the theorem's conclusion. The rates are derived by controlling the empirical process with covering-number estimates (Lemmas S.5 and S.6) and with concentration inequalities for linear processes (Lemmas S.1 and S.3) and martingales (Lemma S.4). Some of these lemmas are cited from prior work by the corresponding author (Zheng 2025; Huang et al. 2024), but they are generic spectral, Hanson-Wright, and martingale bounds whose stated assumptions do not contain the tensor-regression error bound; they are standard probabilistic tools, so this is not load-bearing self-citation in the circularity sense. The empirical claims are out-of-sample rolling forecasts on held-out periods (Section S4.1: Q1-2022 through Q4-2022; Section S4.2: year 2017), not in-sample fitted constants relabeled as predictions. The paper itself flags the main limitations in Section 6: rank and sparsity parameters are selected by time-series cross-validation without a consistency theory, and the alternating-minimization algorithm's convergence is not established. A further validity gap, but not a circularity, is that in the macroeconomic application T=220, R=40, and sum_d p_d + sum_d q_d = 294, so the theorem's stated sample-size condition T >= C R (sum_d p_d + sum_d q_d) log N is not satisfied; the empirical results therefore lie outside the theorem's formal regime. None of these considerations make the theoretical derivation equivalent to its inputs.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The paper introduces no new physical or conceptual entities; it applies an existing tensor decomposition (CP) to a new setting. All assumptions are standard domain conditions for high-dimensional time series and tensor regression. The free parameters are the tuning parameters R and s_d, which are selected in practice via cross-validation.

free parameters (2)
  • CP rank R = R=40 for macro data, R=4 for air pollution data (selected by cross-validation)
    The rank is assumed known in the theory (Assumption 1) but chosen by time-series cross-validation in the empirical sections. The error bounds depend on R.
  • Sparsity levels s_d = s=(4,3,3,4,4,12) for air pollution data
    Sparsity levels are assumed known in Assumption 4 and selected by cross-validation in practice. The sparse estimator's rate depends on s = prod s_d.
assumptions (5)
  • domain assumption Assumption 1: B* has a unique CP decomposition of rank R up to permutation (Kruskal-rank condition).
    Ensures identifiability of the parameterization; without it the estimator may not converge to B*.
  • domain assumption Assumption 2: Innovations are sub-Gaussian with independent coordinates and are martingale differences with respect to the predictor filtration.
    Enables martingale concentration bounds used in the proofs of Theorems 1 and 2.
  • domain assumption Assumption 3: The predictor process is stationary, has an infinite-order moving average representation with summable coefficients, and the innovation process is sub-Gaussian with dimension q_epsilon <= q_x.
    Provides the restricted strong convexity of the loss function via a time-series Hanson-Wright inequality.
  • domain assumption Assumption 4: The true loading vectors are sparse with ||beta*_{r,d}||_0 <= s_d.
    Defines the parameter space for the sparse estimator and drives the improved rate in Theorem 2.
  • standard math Background results: Hanson-Wright inequality for stationary time series (Lemma S.3), martingale concentration (Lemma S.4), covering number bounds (Lemmas S.5, S.6), and the spectral bounds of the linear process (Lemma S.1).
    These lemmas are cited from prior literature (Zheng 2025, Huang et al. 2024) and used as technical tools; they are not proven in this paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Tensor Stochastic Regression for High-dimensional Time Series via CP Decomposition." pith.science (2026). https://pith.science/paper/F3IM4VBL

@misc{pith2026250606919,
  author       = {Pith},
  title        = {Pith review of: Tensor Stochastic Regression for High-dimensional Time Series via CP Decomposition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/F3IM4VBL}},
  note         = {Machine review of arXiv:2506.06919}
}
read the original abstract

As tensor-valued data become increasingly common in time series analysis, there is a growing need for flexible and interpretable models that can handle high-dimensional predictors and responses across multiple modes. We propose a unified framework for high-dimensional tensor stochastic regression based on CANDECOMP/PARAFAC (CP) decomposition, which encompasses vector, matrix, and tensor responses and predictors as special cases. Tensor autoregression naturally arises as a special case within this framework. By leveraging CP decomposition, the proposed models interpret the interactive roles of any two distinct tensor modes, enabling dynamic modeling of input-output mechanisms. We develop both CP low-rank and sparse CP low-rank estimators, establish their non-asymptotic error bounds, and propose an efficient alternating minimization algorithm for estimation. Simulation studies confirm the theoretical properties and demonstrate the computational advantage. Applications to mixed-frequency macroeconomic data and spatio-temporal air pollution data reveal interpretable low-dimensional structures and meaningful dynamic dependencies.

Figures

Figures reproduced from arXiv: 2506.06919 by the authors.

Figure 1
Figure 1. Estimation error }Br ´ B}F against T in the non-sparse case for different p and R. independently from a stationary AR(1) process Xi,j,k,t “ αi,j,kXi,j,k,t´1 ` ǫi,j,k,t, with ǫi,j,k,t i.i.d. „ Np0, 1q. • DGP 2 (Tensor autoregression): Yt “ xB, Yt´1y ` Et , where Yt P R p1ˆp2ˆp3 . For both DGPs, vecpEtq i.i.d. „ Np0, Iq, and we set pd “ qd :“ p for 1 ď d ď 3. For DGP 1, the coefficients αi,j,k are drawn from Unifp´1, … view at source ↗
Figure 2
Figure 2. Estimation error }Bp ´ B}F against T in the sparse case for different s0 and R. 100 120 140 160 1.0 1.5 2.0 2.5 3.0 3.5 4.0 || ~ B − B|| T CP p = 5 CP p = 7 Tucker p = 5 Tucker p = 7 100 120 140 160 1.5 2.0 2.5 3.0 3.5 4.0 || ^B − B|| T CP s0 = 3 CP s0 = 4 Tucker s0 = 3 Tucker s0 = 4 [PITH_FULL_IMAGE:figures/full_fig_p020_2.png] view at source ↗
Figure 3
Figure 3. Estimation errors of the proposed estimators in the non- [PITH_FULL_IMAGE:figures/full_fig_p020_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Heatmaps for estimates of βr,1 ˝ βr,2 (matrices at top) and β J r,3 (row vectors at bottom) for macroeconomic data, for r “ 1 (left panel) and 2 (right panel), where r indexes ωr in descending order. The rows and columns of βr,1 ˝βr,2 correspond to quarterly responses …
Figure 5
Figure 5. Figure 5: Heatmaps for estimates of βr,d1 ˝ βr,d2 for Taiwan air pollution data with r “ 1 and 2. The rows and columns of βr,d1 ˝ βr,d2 correspond to the d1th and d2th modes of B, respectively. The axis for monitoring stations is partitioned into three regions (from top to botto…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

50 extracted references · 49 canonical work pages

  1. [1]

    Babii, A., Ghysels, E., and Pan, J. (2024). Tensor PCA for factor models. Available at SSRN 4791809

    work page 2024

  2. [2]

    Babii, A., Ghysels, E., and Striaukas, J. (2022). Machine Learning Time Series Regressions With an Application to Nowcasting . Journal of Business and Economic Statistics , 40:1094--1106

    work page 2022

  3. [3]

    and Wang, P

    Bai, J. and Wang, P. (2016). Econometric analysis of large factor models. Annual Review of Economics , 8:53--80

    work page 2016

  4. [4]

    Barigozzi, M., He, Y., Li, L., and Trapani, L. (2022). Statistical inference for large-dimensional tensor factor model by iterative projections. arXiv preprint arXiv:2206.09800

    work page Pith review arXiv 2022

  5. [5]

    and Michailidis, G

    Basu, S. and Michailidis, G. (2015). Regularized estimation in sparse high-dimensional time series models. The Annals of Statistics , 43:1535--1567

    work page 2015

  6. [6]

    Bi, X., Tang, X., Yuan, Y., Zhang, Y., and Qu, A. (2021). Tensors in statistics. Annual Review of Statistics and Its Application , 8:345--368

    work page 2021

  7. [7]

    Chan, J. C. C. and Qi, Y. (2024). Large bayesian tensor vars with stochastic volatility. arXiv preprint arXiv:2409.16132

    work page Pith review arXiv 2024

  8. [8]

    Chang, J., He, J., Yang, L., and Yao, Q. (2023). Modelling matrix time series via a tensor cp-decomposition. Journal of the Royal Statistical Society Series B: Statistical Methodology , 85:127--148

    work page 2023

Show all 50 references
  1. [9]

    Chen, B., Han, Y., and Yu, Q. (2024). Estimation and inference for CP tensor factor models. arXiv preprint arXiv:2406.17278

    2024

  2. [10]

    Chen, R., Xiao, H., and Yang, D. (2021a). Autoregressive models for matrix-valued time series. Journal of Econometrics , 222:539--560

    2021

  3. [11]

    Chen, R., Yang, D., and Zhang, C.-H. (2022). Factor models for high-dimensional tensor time series. Journal of the American Statistical Association , 117:94--116

    2022

  4. [12]

    and Lam, C

    Chen, W. and Lam, C. (2024). Rank and factor loadings estimation in time series tensor factor model by pre-averaging. The Annals of Statistics , 52:364--391

    2024

  5. [13]

    Chen, Y.-L., Kolar, M., and Tsay, R. S. (2021b). Tensor canonical correlation analysis with convergence and statistical guarantees. Journal of Computational and Graphical Statistics , 30:728--744. Data available at https://github.com/youlinchen/TCCA

    2021

  6. [14]

    Feng, L., Bi, X., and Zhang, H. (2021). Brain regions identified as being associated with verbal reasoning through the use of imaging regression via internal variation. Journal of the American Statistical Association , 116(533):144--158

    2021

  7. [15]

    Foroni, C., Marcellino, M., and Schumacher, C. (2015). Unrestricted mixed data sampling (midas): Midas regressions with unrestricted lag polynomials. Journal of the Royal Statistical Society. Series A (Statistics in Society) , 178:57--82

    2015

  8. [16]

    and Marcellino, M

    Foroni, C. and Marcellino, M. G. (2013). A survey of econometric methods for mixed-frequency data. Available at SSRN 2268912

    2013

  9. [17]

    Ghysels, E., Sinko, A., and Valkanov, R. (2006). MIDAS regressions: Further results and new directions. Econometric Reviews , 26:53--90

    2006

  10. [18]

    Han, Y., Yang, D., Zhang, C.-H., and Chen, R. (2024). CP factor model for dynamic tensors . Journal of the Royal Statistical Society Series B: Statistical Methodology , 86:1383–1413

    2024

  11. [19]

    Huang, F., Lu, K., Zheng, Y., and Li, G. (2025). Supervised factor modeling for high-dimensional linear time series. Journal of Econometrics , 249:105995

    2025

  12. [20]

    Kolda, T. G. and Bader, B. W. (2009). Tensor decompositions and applications. SIAM Review , 51:455--500

    2009

  13. [21]

    Kruskal, J. B. (1976). More factors than subjects, tests and treatments: an indeterminacy theorem for canonical decomposition and individual differences scaling. Psychometrika , 41:281--293

    1976

  14. [22]

    Liu, J., Zhu, C., Long, Z., and Liu, Y. (2021). Tensor regression. Foundations and Trends® in Machine Learning , 14:379--565

    2021

  15. [23]

    Lock, E. F. (2018). Tensor-on-tensor regression. Journal of Computational and Graphical Statistics , 27:638--647

    2018

  16. [24]

    and Zhang, A

    Luo, Y. and Zhang, A. R. (2024). Tensor-on-tensor regression: Riemannian optimization, over-parameterization, statistical-computational gap, and their interplay. Annals of Statistics . to appear

    2024

  17. [25]

    L\" u tkepohl, H. (2005). New Introduction to Multiple Time Series Analysis . Springer Science & Business Media

    2005

  18. [26]

    McCracken, M. W. and Ng, S. (2016). FRED-MD : A monthly database for macroeconomic research. Journal of Business & Economic Statistics , 34:574--589

    2016

  19. [27]

    McCracken, M. W. and Ng, S. (2021). FRED-QD: A quarterly database for macroeconomic research . Federal Reserve Bank of St. Louis Review , 103:1--44

    2021

  20. [28]

    Raskutti, G., Yuan, M., and Chen, H. (2019). Convex regularization for high-dimensional multi-response tensor regression. Annals of Statistics , 47:1554--1584

    2019

  21. [29]

    Rogers, M., Li, L., and Russell, S. J. (2013). Multilinear dynamical systems for tensor time series. Advances in Neural Information Processing Systems , 26

    2013

  22. [30]

    Samadi, S. Y. and Billard, L. (2024). On a matrix-valued autoregressive model. Journal of Time Series Analysis . to appear

    2024

  23. [31]

    Sidiropoulos, N. D. and Bro, R. (2000). On the uniqueness of multilinear decomposition of n-way arrays. Journal of Chemometrics , 14:229--239

    2000

  24. [32]

    Sun, W. W. and Li, L. (2017). STORE: Sparse tensor response regression and neuroimaging analysis . Journal of Machine Learning Research , 18:1--37

    2017

  25. [33]

    Ten Berge, J. M. and Sidiropoulos, N. D. (2002). On uniqueness in CANDECOMP/PARAFAC . Psychometrika , 67:399--409

    2002

  26. [34]

    Tsay, R. S. (2024). Matrix-variate time series analysis: A brief review and some new developments. International Statistical Review , 92:246--262

    2024

  27. [35]

    P., Reinsel, G

    Velu, R. P., Reinsel, G. C., and Wichern, D. W. (1986). Reduced rank models for multiple time series. Biometrika , 73:105--118

    1986

  28. [36]

    Wang, D., Liu, X., and Chen, R. (2019). Factor models for matrix-valued high-dimensional time series. Journal of Econometrics , 208:231--248

    2019

  29. [37]

    Wang, D., Zheng, Y., and Li, G. (2024). High-dimensional low-rank tensor autoregressive time series modeling. Journal of Econometrics , 238(1):105544

    2024

  30. [38]

    Wang, D., Zheng, Y., Lian, H., and Li, G. (2022). High-dimensional vector autoregressive time series modeling via tensor decomposition. Journal of the American Statistical Association , 117(539):1338--1356

    2022

  31. [39]

    Wang, X., Zhu, H., and Initiative, A. D. N. (2017). Generalized scalar-on-image regression models via total variation. Journal of the American Statistical Association , 112(519):1156--1168

    2017

  32. [40]

    and Ng, M

    Zeng, C. and Ng, M. K. (2021). Incremental CP tensor decomposition by alternating minimization method. SIAM Journal on Matrix Analysis and Applications , 42:832--858

    2021

  33. [41]

    Zheng, Y. (2025). An interpretable and efficient infinite-order vector autoregressive model for high-dimensional time series. Journal of the American Statistical Association , 120:212--225

    2025

  34. [42]

    Zhou, H., Li, L., and Zhu, H. (2013). Tensor regression with applications in neuroimaging data analysis. Journal of the American Statistical Association , 108(502):540--552

    2013

  35. [43]

    and Teboulle, M

    Beck, A. and Teboulle, M. (2009). Iterative shrinkage and thresholding for linear inverse problems. SIAM Journal on Imaging Sciences , 2(1):183--202

    2009

  36. [44]

    M., and Chen, Y

    Chi, Y., Lu, Y. M., and Chen, Y. (2019). Nonconvex optimization meets low-rank matrix factorization: An overview. IEEE Transactions on Signal Processing , 67:5239--5269

    2019

  37. [45]

    De Lathauwer, L., De Moor, B., and Vandewalle, J. (2000). A multilinear singular value decomposition. SIAM journal on Matrix Analysis and Applications , 21(4):1253--1278

    2000

  38. [46]

    Huang, F., Lu, K., and Zheng, Y. (2024). SARMA: S calable low-rank high-dimensional autoregressive moving averages via tensor decomposition. ArXiv preprint arXiv:2405.00626

    2024 arXiv

  39. [47]

    J., and Yu, B

    Raskutti, G., Wainwright, M. J., and Yu, B. (2011). Minimax rates of estimation for high-dimensional linear regression over _q -balls. IEEE Transactions on Information Theory , 57(10):6976--6994

    2011

  40. [48]

    W., Lu, J., Liu, H., and Cheng, G

    Sun, W. W., Lu, J., Liu, H., and Cheng, G. (2017). Provable sparse tensor decomposition. Journal of the Royal Statistical Society Series B: Statistical Methodology , 79(3):899--916

    2017

  41. [49]

    Tibshirani, R. (1996). Regression shrinkage and selection via the lasso. Journal of the Royal Statistical Society: Series B , 58:267–288

    1996

  42. [50]

    Wilms, I., Basu, S., Bien, J., and Matteson, D. (2023). Sparse identification and estimation of large-scale vector autoregressive moving averages. Journal of the American Statistical Association , 118:571–582

    2023

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.