Pith. sign in

REVIEW 3 major objections 7 minor 43 references

Low-rank matrix recovery via nonconvex optimization methods with application to errors-in-variables matrix regression

T0 review · 3 major / 7 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 For any stationary point of a nonconvex low-rank estimator, the paper proves a recovery bound that beats the nuclear-norm bound when some singular values are large; the application is errors-in-variables matrix regression.

desk verdict A worthwhile stationary-point extension undone by two proof gaps: a threshold mismatch in the deterministic part and a false spectral-norm identity in the probabilistic part. read the letter →

arxiv 2412.16263 v1 pith:RMKWVSX4 submitted 2024-12-20 math.OC math.STstat.TH

Xin Li , Dongya Wu This is my paper · ORCID
classification math.OCmath.STstat.TH MSC 90C2662J0715A42
keywords nonconvexoptimizationlow-rankmatrixrecoveryregularizationerrors-in-variablesregressionboundsstationarypointsrestrictedstrongconvexitySCADandMCPpenalties
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 proves that a nonconvex penalty on the singular values of a low-rank matrix estimator still guarantees recovery at every stationary point, not just at global minima, provided the loss is strongly convex enough to outweigh the concavity of the penalty. The proof's error bound splits the true singular values into a large-value set $J_1$, where the penalty has flattened and therefore no longer distorts the estimate, and a small-value set $J_2$, where only a mild $\lambda$-proportional term remains. Consequently, when some singular values exceed the threshold $\nu$ built into regularizers such as SCAD or MCP, the bound is tighter than the nuclear-norm bound, with the best case giving a rate of order $\sqrt{r\log r/N}$ instead of $\sqrt{r\tilde d/N}$. The paper also verifies the required conditions for errors-in-variables matrix regression with additive noise and missing data, where the empirical loss is itself nonconvex.

What carries the argument

The central object is the flat spectral regularizer $P_\lambda(\Theta)=\sum_j p_\lambda(\sigma_j(\Theta))$ with $p_\lambda=q_\lambda+\lambda|\cdot|$, where $q_\lambda$ is concave, flat in the sense that $p_\lambda'(t)=0$ for $t\ge\nu$, and has curvature bounded by $\mu$ (SCAD and MCP are named instances). The proof also uses the split $J_1=\{j:\sigma_j(\Theta^*)\ge\nu\}$, $J_2=\{j:0<\sigma_j(\Theta^*)<\nu\}$ together with the subspaces $A_{J_1},A_{J_2}$ and $B_J$ built from the singular vectors of $\Theta^*$. The mechanism is that the stationary-point inequality, combined with local strong convexity, first forces the error into the cone $\{\Delta:\|\Delta_{T^c}\|_*\le 7\|\Delta_T\|_*\}$; restricted strong convexity then applies, and a carefully chosen subgradient $G^*=U^*V^{*\top}-\lambda^{-1}\Pi_{B_J}(\nabla L_N(\Theta^*))$ cancels the $B_J$ component. What remains is the projected gradient on $A_{J_1}$, where the penalty is flat, plus a $\lambda\sqrt{r_2}$ term from the small singular values.

What would settle it

Set $N=1$ in the additive-noise errors-in-variables model with $\Sigma_x=I$, $\Sigma_w=0$, and nonzero noise $\epsilon$, so that $\nabla L_1(\Theta^*)=\epsilon\,\mathrm{mat}(x)$ for a standard normal design vector $x$. The entrywise-max norm of the vectorized gradient is about $|\epsilon|\sqrt{2\log\tilde d}$ while its operator norm is about $2|\epsilon|\sqrt{\tilde d}$; checking which of these two quantities controls the actual error of a stationary point would settle whether the projected-gradient bound holds at the claimed $\sqrt{\log r_1/N}$ scale.

Watch

Extended reading notes

Core claim

Let $\Theta^*$ be a rank-$r$ parameter matrix and let $\hat\Theta$ be any stationary point of the problem $\min\{L_N(\Theta)+P_\lambda(\Theta): \|\Theta\|_*\le\omega\}$, where $P_\lambda$ is a spectral nonconvex penalty of the form $\sum_j p_\lambda(\sigma_j(\Theta))$ with $p_\lambda=q_\lambda+\lambda|\cdot|$ and $q_\lambda$ concave. Theorem 1 states that if the loss satisfies local strong convexity and restricted strong convexity with parameters $\alpha_1,\alpha_2>0$, the penalty's concavity parameter $\mu$ satisfies $\min\{\alpha_1,2\alpha_2\}>\mu$, and $\lambda\ge 2\max\{\|\nabla L_N(\Theta^*)\|_{\mathrm{op}},4\omega\tau_1\}$, then $\|\hat\Theta-\Theta^*\|_F \le \frac{\sqrt{r_1}}{2\alpha_2-\mu}\|\Pi_{A_{J_1}}(\nabla L_N(\Theta^*))\|_{\mathrm{op}} + \frac{5\sqrt{r_2}}{2(2\alpha_2-\mu)}\lambda$, where $J_1$ indexes nonzero singular values of $\Theta^*$ at least as large as the flatness threshold $\nu$, $J_2$ indexes the remaining ones, and $r_1+r_2=r$. Because the penalty derivative vanishes on $J_1$, the first term sees only a low-rank projection of the gradient rather than the full gradient, and the second term pays only for the small singular values. In the errors-in-variables application the paper proves that these conditions hold with high probability for additive noise and missing data, giving the explicit rates in Corollaries 1 and 2 that improve on the nuclear-norm baseline whenever $r_1>0$.

Load-bearing premise

The load-bearing premise is that the largest 'stretch factor' of the gradient matrix at the true answer, even after projecting onto the large-singular-value direction, is controlled by its largest single entry and so scales like $\sqrt{\log r/N}$; if the true stretch factor is $\sqrt{r_1}$ times larger, as it is for generic dense random matrices, the advertised nonconvex improvement does not follow.

Editorial extensions

If this is right

  • Any algorithm that converges to a stationary point of (5) satisfies the same recovery guarantee, so the theory applies to first-order and local-search methods without requiring a global solution.
  • When at least one singular value of the true matrix exceeds the threshold $\nu$, the Frobenius error rate is smaller than the nuclear-norm rate $O(\sqrt{r\tilde d/N})$ of the comparison methods, because the projected gradient term replaces the full gradient norm.
  • If all $r$ singular values are above the threshold, the second term vanishes and the rate becomes $O(\sqrt{r\log r/N}\,)$ up to constants, with no dependence on the ambient dimension $\tilde d$.
  • For errors-in-variables matrix regression, the same rate holds when covariates are observed with additive Gaussian noise or with entries missing at random, once the sample size satisfies $N\gtrsim \tau \tilde d \log \tilde d$.

Reading between the lines

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

  • Editorial inference: the proof's gradient bounds identify the largest stretch factor of $\nabla L_N(\Theta^*)$ with the largest entry of its vectorized form; for dense Gaussian designs this identity fails by a factor of order $\sqrt{r_1}$ at least, so the advertised dimension-free rate likely needs the projected gradient to be sparse or low-rank in a stronger sense than the paper states.
  • Editorial inference: the same flat-penalty mechanism should transfer to other penalties with a flat region, such as capped-$\ell_1$ or firm-threshold penalties, since only Assumption 1(iv) is used to kill the $A_{J_1}$ term; this is a testable family of extensions.
  • Editorial inference: a numerical check of the rate is direct: simulate the additive-noise model with dense Gaussian covariates, run any algorithm to a stationary point, and compare the empirical Frobenius error to $\sqrt{r\log r/N}$ and to $\sqrt{r\tilde d/N}$; the scaling in $r_1$ versus $\tilde d$ distinguishes the claimed bound from the spectral-norm fallback.
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

3 major / 7 minor

Summary. The paper studies nonconvex regularized low-rank matrix recovery for the estimator (5), where both the loss and the spectral regularizer may be nonconvex. The main deterministic result, Theorem 1, claims a recovery bound for every stationary point under local strong convexity (LSC), restricted strong convexity (RSC), and Assumption 1 on the regularizer; the bound depends on the projection of the gradient onto the subspace associated with large singular values. The paper then applies this framework to errors-in-variables matrix regression with additive noise and missing data, claiming rates of order sqrt(r1 log r1 / N) + sqrt(r2 dtilde log dtilde / N), which are advertised as tighter than the nuclear-norm rate sqrt(r dtilde / N).

Significance. If the main results were correct, the paper would provide a useful stationary-point theory for nonconvex spectral regularization, going beyond existing global-solution analyses, and would offer the first rate guarantees for nonconvex errors-in-variables matrix regression. The paper is also well organized, covers both additive-noise and missing-data settings, and explicitly verifies the abstract regularity conditions using external concentration lemmas. However, the two central advances — the refined deterministic bound and the dimension-free probabilistic improvement — rest on specific technical steps that are currently invalid, so the advertised significance is not established.

major comments (3)
  1. [Section 4, Proposition 1(iii) and Proposition 2(iii)] The proof asserts the equality ||∇L_N(Θ*)||_op = ||Γ̂ vec(Θ*) − Υ̂||_∞. This equality is false in general: the left-hand side is the spectral norm of the matrix whose vectorization is Γ̂ vec(Θ*) − Υ̂, and the spectral norm is not the maximum absolute entry. Lemma 2 of [19] bounds only the maximum absolute entry of a sub-Gaussian random vector, so it cannot deliver the stated operator-norm rates. Consequently the bounds in Corollaries 1 and 2, and the comparison with the nuclear-norm rate in Remark 3, are unsupported. For an isotropic Gaussian design, the operator norm is typically of order φ sqrt(dtilde/N), not φ sqrt(log dtilde/N), and the projected term is at best φ sqrt(r1 log r1/N) on the paper's own entrywise reasoning. Replacing dtilde by log r1 in the first error term is exactly the advertised improvement, so this is a load-bearing error.
  2. [Theorem 1 proof, Eq. (43), and definitions (23)/(40)] The proof after Eq. (43) uses p'_λ(σ_j(Θ*)) = 0 for j in J1, citing Assumption 1(iv). Assumption 1(iv) gives p'_λ(t) = 0 only for t ≥ ν, while J1 is defined in (23) and (40) by σ_j(Θ*) ≥ μ. No relation between μ and ν is assumed; in the SCAD and MCP examples ν is a singular-value threshold (aλ or bλ) whereas μ is a curvature constant (1/(a−1) or 1/b). Thus the displayed step is not justified as written. The theorem needs either J1 defined with the threshold ν or an explicit assumption relating μ and ν. Since r1 and r2 enter the final bound, this gap affects the main deterministic claim.
  3. [Lemma 3, Eqs. (19a)–(19b)] The proof reduces the matrix inequality to univariate inequalities for the paired singular values (σ_j, σ'_j). This reduction is not justified: ∇Qλ(Θ) = U diag(q'_λ(σ)) V^⊤, and the trace inner product with Θ−Θ' depends on the alignment of the singular vectors of Θ and Θ'; it is not equal to ∑_j (q'_λ(σ_j) − q'_λ(σ'_j))(σ_j − σ'_j). Since (19a) and (19b) are used in the proofs of Lemma 5 and Theorem 1, a correct proof or a precise citation for the spectral-function inequality is needed.
minor comments (7)
  1. [Section 4, after Eq. (56)] The definition '˜d = max{d2, d2}' should read '˜d = max{d1, d2}'.
  2. [Section 4, Eq. (59)] In the definition of φadd, the expression (||Σx||_op + ||Σx||_op) is presumably meant to be (||Σx||_op + ||Σw||_op); otherwise the additive-noise parameter is not symmetric in the noise.
  3. [Proposition 1(i) proof] The proof fixes Δ with Θ* + Δ ∈ S, but the set S has not been defined.
  4. [Proposition 2] There are two items labeled (ii); the second one (the probabilistic bounds on ∇L_N(Θ*) and its projection) should be labeled (iii).
  5. [Lemma 5 proof] The proof begins 'By the RSC condition (10)', but Eq. (10) is the LSC condition; the label appears to be a typographical error.
  6. [Corollaries 1 and 2] The sample-size condition 'N ≥ 4c0 τadd ˜d log ˜d / N' places N on both sides of the inequality; this should be corrected to the intended lower bound N ≳ τadd ˜d log ˜d.
  7. [Throughout] There are several duplicated words and phrases, e.g., 'advantage is advantage is due to' in Remark 1(ii) and 'then then it holds' in Corollaries 1 and 2.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Theorem 1 is derived from stated LSC/RSC and regularizer assumptions, not fitted to its target bound; self-citations are non-load-bearing.

full rationale

No circular step is present in the derivation chain. Theorem 1's recovery bound is obtained from the LSC/RSC inequalities (10)-(11), Assumption 1 on the regularizer, and the stationary-point condition (26); no parameter is fitted to the quantity being bounded, and the projected-gradient term Π_AJ1(∇L_N(Θ*)) appears on the right-hand side only after a subgradient choice that makes the AJ1 component of ∇Pλ(Θ*) vanish. The key technical inputs are external: [19, Lemmas 1-4] for sub-Gaussian concentration, [30,40] for singular-value inequalities, [18] for pλ bounds, and [25,27] for decompositions of Δ. The paper's self-citations [15,16] appear only as related work and are not used to close any proof. There are genuine correctness concerns that are not circularity: Proposition 1(iii) and Proposition 2(iii) assert ||∇L_N(Θ*)||_op = ||Γ̂ vec(Θ*) − Υ̂||_∞, which is false for dense matrices and is not what [19, Lemma 2] bounds; and Theorem 1's proof uses σ_j(Θ*) ≥ ν to get p'_λ=0 for j∈J1, while J1 was defined by σ_j(Θ*) ≥ μ, so a μ-ν relation is missing. These are mathematical-support gaps, not reductions of the claimed bound to its own inputs.

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

The proof relies on the low-rank assumption, the LSC and RSC regularity conditions on the loss, and Assumption 1 on the regularizer, plus standard convex-analysis facts about spectral functions and nuclear-norm subgradients. No constants are fitted to data in the derivation; the tuning parameters lambda and omega are part of the estimator definition, and model primitives such as Sigma_x, Sigma_w, and sigma_epsilon are inputs to the probabilistic verification. No new entities are introduced beyond the estimator itself.

assumptions (6)
  • domain assumption The true parameter Theta* is exactly low-rank with rank(Theta*) <= r and lies in the feasible nuclear-norm ball Omega.
    Used in Eq. (4) and Lemma 5 to localize the error and to bound ||Delta||_* <= 2*omega.
  • domain assumption The loss L_N satisfies local strong convexity (LSC, Eq. 10) with alpha1 > mu and restricted strong convexity (RSC, Eq. 11) with 2*alpha2 > mu.
    These are the core regularity conditions in Theorem 1; they are verified probabilistically in Section 4, though that verification is flawed.
  • domain assumption The regularizer P_lambda satisfies Assumption 1: concave, symmetric, flat beyond nu, and with mu-weak convexity of the concave part q_lambda.
    Assumption 1 supplies the subadditivity, flat-penalty, and curvature bounds used in Lemmas 2 through 5. SCAD and MCP satisfy it.
  • standard math Standard singular value inequalities for concave functions (Lemma 1), cited from Rotfel'd and Yue and Anthony.
    Used in Lemma 2 to prove subadditivity of the regularizer P_lambda.
  • standard math The weak convexity of the spectral function Q_lambda follows from entrywise inequalities on singular values and Nesterov's convex analysis results (Lemma 3).
    The proof given is abbreviated and ignores subspace rotations, but the underlying statement is plausible via convex spectral function arguments.
  • standard math The subdifferential of the nuclear norm at Theta* = U* D* V*^T is U* V*^T + W with ||W||_op <= 1 and W orthogonal to the row and column spaces.
    Used in Eq. (42) to construct a favorable subgradient that cancels the projection onto A_J1 in the proof of Theorem 1.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Low-rank matrix recovery via nonconvex optimization methods with application to errors-in-variables matrix regression." pith.science (2026). https://pith.science/paper/RMKWVSX4

@misc{pith2026241216263,
  author       = {Pith},
  title        = {Pith review of: Low-rank matrix recovery via nonconvex optimization methods with application to errors-in-variables matrix regression},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RMKWVSX4}},
  note         = {Machine review of arXiv:2412.16263}
}
read the original abstract

We consider the nonconvex regularized method for low-rank matrix recovery. Under the assumption on the singular values of the parameter matrix, we provide the recovery bound for any stationary point of the nonconvex method by virtue of regularity conditions on the nonconvex loss function and the regularizer. This recovery bound can be much tighter than that of the convex nuclear norm regularized method when some of the singular values are larger than a threshold defined by the nonconvex regularizer. In addition, we consider the errors-in-variables matrix regression as an application of the nonconvex optimization method. Probabilistic consequences and the advantage of the nonoconvex method are demonstrated through verifying the regularity conditions for specific models with additive noise and missing data.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 43 canonical work pages

  1. [19]

    Loh and M

    P.-L. Loh and M. J. Wainwright. High-dimensional regression with noisy and missing data: Provable guarantees with nonconvexity. Ann. Stat. , 40(3):1637–1664, 2012

    work page 2012

  2. [1]

    Agarwal, S

    A. Agarwal, S. Negahban, and M. J. Wainwright. Fast global conv ergence of gradient methods for high-dimensional statistical recovery. Ann. Stat. , 40(5):2452–2482, 2012

    work page 2012

  3. [2]

    Alquier, C

    P. Alquier, C. Butucea, M. Hebiri, K. Meziani, and T. Morimae. Rank -penalized estima- tion of a quantum system. Phys. Rev. A , 88(3):032113, 2013

    work page 2013

  4. [3]

    Becker, E

    S. Becker, E. J. Cand` es, and M. C. Grant. Templates for conv ex cone problems with applications to sparse signal recovery. Math. Program. Comput. , 3:165–218, 2011

    work page 2011

  5. [4]

    Breheny and J

    P. Breheny and J. Huang. Coordinate descent algorithms for no nconvex penalized re- gression, with applications to biological feature selection. Ann. Appl. Stat. , 5(1):232–253, 2011

    work page 2011

  6. [5]

    J.-F. Cai, E. J. Cand` es, and Z. W. Shen. A singular value thresho lding algorithm for matrix completion. SIAM J. Optim. , 20(4):1956–1982, 2010

    work page 1956

  7. [6]

    E. J. Cand` es and Y. Plan. Tight oracle inequalities for low-rank ma trix recovery from a minimal number of noisy random measurements. IEEE Trans. Inf. Theory , 57(4):2342– 2359, 2011

    work page 2011

  8. [7]

    E. J. Cand` es and T. Tao. The power of convex relaxation: Near -optimal matrix comple- tion. IEEE Trans. Inf. Theory , 56(5):2053–2080, 2010

    work page 2010

Show all 43 references
  1. [8]

    R. J. Carroll, D. Ruppert, C. M. Crainiceanu, and L. A. Stefansk i. Measurement error in nonlinear models: A modern perspective . CRC Press, Boca Raton, 2006

    2006

  2. [9]

    Datta and H

    A. Datta and H. Zou. CoCoLasso for high-dimensional error-in- variables regression. Ann. Stat., 45(6):2400–2426, 2017

    2017

  3. [10]

    J. Q. Fan and R. Z. Li. Variable selection via nonconcave penalized likelihood and its oracle properties. J. Am. Stat. Assoc. , 96(456):1348–1360, 2001

    2001

  4. [11]

    Fazel, H

    M. Fazel, H. A. Hindi, and S. P. Boyd. A rank minimization heuristic w ith application to minimum order system approximation. Proceedings of the 2001 American Control Conference. (Cat. No.01CH37148) , 6:4734–4739, 2001. 20

    2001

  5. [12]

    H. Gui, J. W. Han, and Q. Q. Gu. Towards faster rates and orac le property for low-rank matrix estimation. In Proc. ICML, pages 2300–2309, 2015

    2015

  6. [13]

    Y. Hu, D. B. Zhang, J. P. Ye, and X. L. Li. Fast and accurate ma trix completion via truncated nuclear norm regularization. IEEE Trans. Pattern Anal. Machine Intell. , (9):35, 2013

    2013

  7. [14]

    Lee and Y

    K. Lee and Y. Bresler. ADMiRA: Atomic decomposition for minimum r ank approxima- tion. IEEE Trans. Inf. Theory , 56:4402–4416, 2010

    2010

  8. [15]

    Li and D

    X. Li and D. Y. Wu. Low-rank matrix estimation via nonconvex op timization methods in multi-response errors-in-variables regression. J. Global Optim. , 2023

    2023

  9. [16]

    Li and D

    X. Li and D. Y. Wu. Low-rank matrix estimation via nonconvex op timization methods in multi-response errors-in-variables regression. J. Glob. Optim. , 88:79–114, 2024

    2024

  10. [17]

    X. Li, D. Y. Wu, C. Li, J. H. Wang, and J.-C. Yao. Sparse recover y via nonconvex regularized M-estimators over ℓq-balls. Comput. Stat. Data Anal. , 152:107047, 2020

    2020

  11. [18]

    P.-L. Loh. Local optima of nonconvex regularized M-estimator s. Dept. Elect. Eng. Com- put. Sci., UC Berkeley, Berkeley, 2013

    2013

  12. [20]

    Loh and M

    P.-L. Loh and M. J. Wainwright. Supplementary material: High-dim ensional regression with noisy and missing data: Provable guarantees with nonconvexity . Ann. Stat. , 2012

    2012

  13. [21]

    C. Y. Lu, J. H. Tang, S. C. Yan, and Z. C. Lin. Generalized nonco nvex nonsmooth low-rank minimization. In Proc. IEEE CVPR , pages 4130–4137, 2014

    2014

  14. [22]

    Mohan and M

    K. Mohan and M. Fazel. Iterative reweighted algorithms for mat rix rank minimization. J. Mach. Learn. Res. , 13(1):3441–3473, 2012

    2012

  15. [23]

    B. K. Natarajan. Sparse approximate solutions to linear syste ms. SIAM J. Comput. , 24(2):227–234, 1995

    1995

  16. [24]

    Negahban, P

    S. Negahban, P. Ravikumar, M. J. Wainwright, and B. Yu. A unifie d framework for high-dimensional analysis of M-estimators with decomposable regula rizers. Stat. Sci. , 27(4):538–557, 2012

    2012

  17. [25]

    Negahban and M

    S. Negahban and M. J. Wainwright. Estimation of (near) low-ran k matrices with noise and high-dimensional scaling. Ann. Stat. , 39(2):1069–1097, 2011

    2011

  18. [26]

    Nesterov

    Y. Nesterov. Introductory lectures on convex optimization: A basic cour se, volume 87, chapter 2, pages 56,61. Springer Science & Business Media, Berlin, 2 013

  19. [27]

    Recht, M

    B. Recht, M. Fazel, and P. A. Parrilo. Guaranteed minimum-rank solutions of linear matrix equations via nuclear norm minimization. SIAM Rev. , 52(3):471–501, 2010

    2010

  20. [28]

    Rohde and A

    A. Rohde and A. B. Tsybakov. Estimation of high-dimensional low -rank matrices. Ann. Stat., 39(2):887–930, 2011. 21

    2011

  21. [29]

    Rosenbaum and A

    M. Rosenbaum and A. B. Tsybakov. Sparse recovery under ma trix uncertainty. Ann. Stat., 38(5):2620–2651, 2010

    2010

  22. [30]

    S. Yu. Rotfel’d. Remarks on the singular numbers of a sum of com pletely continuous operators. Functional Analysis and Its Applications , 1:252–253, 1967

    1967

  23. [31]

    Sagan and J

    A. Sagan and J. E. Mitchell. Low-rank factorization for rank min imization with nonconvex regularizers. Comput. Optim. Appl. , 79:273–300, 2021

    2021

  24. [32]

    Sørensen, A

    Ø. Sørensen, A. Frigessi, and M. Thoresen. Measurement err or in LASSO: Impact and likelihood bias correction. Stat. Sinica , pages 809–829, 2015

    2015

  25. [33]

    M. J. Wainwright. Structured regularizers for high-dimensiona l problems: Statistical and computational issues. Annu. Rev. Stat. Appl. , 1:233–253, 2014

    2014

  26. [34]

    H. Wang, F. Zhang, Q. Wu, Y. H. Hu, and Y. M. Shi. Nonconvex an d nonsmooth sparse optimization via adaptively iterative reweighted methods. J. Global Optim. , 81:717–748, 2021

    2021

  27. [35]

    Z. R. Wang, H. Liu, and T. Zhang. Optimal computational and st atistical rates of convergence for sparse nonconvex learning problems. Ann. Stat. , 42(6):2164, 2014

    2014

  28. [36]

    J. Wu, Z. M. Zheng, Y. Li, and Y. Zhang. Scalable interpretable le arning for multi- response error-in-variables regression. J. Multivar. Anal. , page 104644, 2020

    2020

  29. [37]

    X. Wu, X. Zhang, N. Wang, and Y. Cen. Joint sparse and low-ran k multi-task learning with extended multi-attribute profile for hyperspectral target d etection. Remote. Sens. , 11:150, 2019

    2019

  30. [38]

    Q. M. Yao, J. T. Kwok, T. F. Wang, and T.-Y. Liu. Large-scale low -rank matrix learning with nonconvex regularizers. IEEE Trans. Pattern Anal. Machine Intell. , 41:2628–2643, 2017

    2017

  31. [39]

    Q. M. Yao, J. T. Kwok, and L. W. Zhong. Fast low-rank matrix lea rning with nonconvex regularization. In Proc. IEEE ICDM , pages 539–548, 2015

    2015

  32. [40]

    Yue and M.-C

    M.-C. Yue and M.-C. S. Anthony. A perturbation inequality for co ncave functions of singular values and its applications in low-rank matrix recovery. Appl Comput. Harmon. A., 40:396–416, 2016

    2016

  33. [41]

    C.-H. Zhang. Nearly unbiased variable selection under minimax con cave penalty. Ann. Stat., 38(2):894–942, 2010

    2010

  34. [42]

    Zhang and J

    C.-H. Zhang and J. Huang. The sparsity and bias of the Lasso se lection in high- dimensional linear regression. Ann. Stat. , 36(4):1567–1594, 2008

    2008

  35. [43]

    Zhou and L

    H. Zhou and L. X. Li. Regularized matrix regression. J. Royal Stat. Soc. B , 76(2):463–483, 2014. 22

    2014

Pith tools

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