arxiv: 2604.20075 · v1 · submitted 2026-04-22 · 💻 cs.IT · math.IT

Recognition: unknown

Robust Uniform Recovery of Structured Signals from Nonlinear Observations

Pedro Abdalla , Radu Balan , Junren Chen

Authors on Pith no claims yet

Pith reviewed 2026-05-09 23:47 UTC · model grok-4.3

classification 💻 cs.IT math.IT

keywords uniform recoverynonlinear observationssingle-index modelprojected gradient descentRAICpiecewise Lipschitz link functionssparse recovery

0 comments

The pith

RAIC enables uniform recovery of structured signals from nonlinear observations via projected gradient descent.

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

The paper establishes that the restricted approximate invertibility condition (RAIC) on loss gradients provides a unified approach to uniform recovery of structured signals from nonlinear observations. Projected gradient descent achieves this recovery whenever its gradients obey RAIC uniformly over the signal class. For the Gaussian single-index model with piecewise Lipschitz link functions, uniform RAIC for the scaled l2 loss gradient is proved via a covering argument, generalizing prior nonuniform guarantees with only modest extra error terms that prove negligible in sparse and 1-bit cases.

Core claim

The restricted approximate invertibility condition (RAIC) on the gradients of the loss allows projected gradient descent to achieve uniform recovery for structured signals from nonlinear measurements. In the Gaussian single-index model under piecewise Lipschitz links, uniform RAIC is verified via covering, producing error rates of the same order as nonuniform guarantees in sparse recovery settings, and without log loss for 1-bit quantization via VC dimension.

What carries the argument

The restricted approximate invertibility condition (RAIC), a uniform property that the loss gradient approximately inverts on the structured signal class, established via covering or VC arguments.

If this is right

Uniform recovery guarantees hold for Gaussian single-index models with piecewise Lipschitz nonlinearities.
Error bounds for iterative hard thresholding and l1-based projected gradient descent match nonuniform rates up to log factors.
Uniform recovery error for 1-bit quantization with iterative hard thresholding matches nonuniform rates with no log factor penalty.
Robustness to noise and corruption is handled by bounding one additional random process capturing gradient mismatch.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Covering arguments may control the uniformity cost across many nonlinear observation models without major degradation.
The RAIC unification could extend to other structured classes or observation models beyond single-index.
Empirical tests on data with piecewise Lipschitz links could check whether the uniform guarantees hold in practice.

Load-bearing premise

The gradient of the loss satisfies the restricted approximate invertibility condition uniformly for all signals in the structured class.

What would settle it

A piecewise Lipschitz link function and signal class where uniform RAIC holds for the gradient but projected gradient descent fails to recover at least one signal from the nonlinear observations.

Figures

Figures reproduced from arXiv: 2604.20075 by Junren Chen, Pedro Abdalla, Radu Balan.

**Figure 2.** Figure 2: The graph of m1/2 (v) = v − ⌊v + 1 2 ⌋. recovery settings that correspond to Remark 4.7 and Remark 4.8: (X , K) = (Σn,∗ k , Σ n k ) (50) and (X , K) = (Σn,∗ k ∩ {u : ∥u∥1 = c∗ √ k}, B n 1 (c∗ √ k)), c∗ ∈ (0, 1). (51) In the two settings, our unified framework yields the following statement that improves on Corollary 5 of Genzel and Stollenwerk (2023), where a uniform recovery error over x ∈ X no faster tha… view at source ↗

read the original abstract

While it is well known that the restricted isometry property (RIP) guarantees uniform sparse recovery from noisy linear measurements, uniform recovery of structured signals from nonlinear observations remains much less understood. This paper shows that the restricted approximate invertibility condition (RAIC) provides a unified approach to this end. Particularly, uniform recovery is achieved by projected gradient descent (PGD) with gradients obeying RAIC for all signals. As an application, under a large class of piecewise Lipschitz link functions (possibly discontinuous), we develop a uniform recovery theory for Gaussian single-index model by establishing the uniform RAIC for the gradient of the (scaled) $\ell_2$ loss via a covering argument. The theory generalizes the nonuniform recovery guarantees due to Plan and Vershynin (2016); Oymak and Soltanolkotabi (2017) and exhibits additional error terms that can be interpreted as the cost of uniform recovery. Intriguingly, in the three canonical settings of (a) sparse recovery via PGD with $\ell_0$ projection (i.e., iterative hard thresholding (IHT)), (b) sparse recovery via PGD with $\ell_1$ projection, and (c) recovering approximately sparse signals via PGD with $\ell_1$ projection, the additional error terms are negligible and in turn our uniform recovery error rates are at the same order of existing nonuniform ones, up to log factors. Our results hence improve on Genzel and Stollenwerk (2023). Under the specific nonlinearity of 1-bit quantization, we use a VC dimension argument to show that the uniform recovery error of IHT is at the same order of the nonuniform recovery error, with no loss of log factor. In addition, we show that the robustness of PGD to noise and corruption can be incorporated elegantly by bounding a single additional random process that captures the gradient mismatch.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

RAIC on the gradient gives uniform recovery via PGD for piecewise Lipschitz single-index models, with the extra uniformity terms staying small enough in the three standard sparse settings to match nonuniform rates up to logs.

read the letter

The paper's central move is to isolate RAIC as a sufficient condition on the gradient of the scaled l2 loss so that projected gradient descent recovers structured signals uniformly from nonlinear observations. They then verify that RAIC holds uniformly over the signal class for Gaussian single-index models when the link is piecewise Lipschitz, using covering numbers in general and VC dimension for the 1-bit case. This turns the known nonuniform guarantees into uniform ones without a large penalty in the rates. In the three canonical projection settings (IHT, l1 projection for sparse, and l1 for approximately sparse) the added error terms from uniformity are negligible, so the final bounds sit at the same order as Plan-Vershynin and Oymak-Soltanolkotabi up to logs, and they improve the earlier uniform result of Genzel-Stollenwerk. The robustness extension is handled by bounding one extra mismatch process, which keeps the argument clean. The covering and VC steps look standard and are applied directly to the gradient, with no obvious post-hoc fitting or circularity. The reduction from uniform RAIC to PGD convergence is straightforward and does not hide extra assumptions. A minor limitation is that the work stays inside the Gaussian single-index model, so the constants are not optimized and extensions to other measurement distributions or more general nonlinearities are left open. The log factors are explicit and the paper does not claim they vanish. This is useful reading for people working on theoretical compressed sensing and statistical signal recovery who need uniform guarantees. The claims are grounded in reproducible tools and the comparisons to prior nonuniform results are concrete, so the paper deserves a serious referee. I would send it to review.

Referee Report

2 major / 3 minor

Summary. The paper claims that the restricted approximate invertibility condition (RAIC) on the gradients of the scaled ℓ₂ loss enables projected gradient descent (PGD) to achieve uniform recovery of structured signals from nonlinear observations. For the Gaussian single-index model with piecewise Lipschitz link functions (possibly discontinuous), uniform RAIC is established via covering arguments, yielding recovery error rates matching nonuniform baselines up to log factors in the settings of sparse recovery via IHT (ℓ₀ projection), sparse recovery via ℓ₁ projection, and approximately sparse signals via ℓ₁ projection. For 1-bit quantization a VC-dimension argument shows no log-factor loss. Robustness to noise and corruption follows from bounding one additional gradient-mismatch random process. The results generalize the nonuniform guarantees of Plan-Vershynin (2016) and Oymak-Soltanolkotabi (2017) and improve on Genzel-Stollenwerk (2023).

Significance. If the derivations hold, the work supplies a unified, practical route to uniform recovery guarantees in nonlinear compressed sensing, which is stronger and more applicable than nonuniform results. The demonstration that the uniformity overhead is only logarithmic (or absent for 1-bit) in the three canonical projection settings is a concrete advance. The tailored covering and VC arguments for general piecewise-Lipschitz links, together with the clean single-process robustness extension, are technical strengths that extend the literature.

major comments (2)

[main recovery theorem and covering argument for uniform RAIC] The central claim that the extra uniformity terms remain negligible (and do not change the order of the recovery error) in the three canonical settings rests on the explicit dependence of the RAIC constants on the covering number of the structured class; this dependence must be tracked through the PGD convergence analysis to confirm negligibility.
[robustness extension paragraph] The robustness statement is presented as following directly from a single additional random-process bound; the precise assumptions on the noise and corruption (e.g., sub-Gaussian tails, bounded support) that make this bound uniform over the signal class should be stated explicitly so that the extension does not introduce hidden structural requirements.

minor comments (3)

[Introduction] The definition of RAIC and the precise scaling of the ℓ₂ loss should be recalled in the introduction for readers who skip the preliminaries.
[Section on single-index model] Notation for the piecewise-Lipschitz link function (number of pieces, discontinuity locations) is used throughout but is only defined later; a compact summary table or early display would improve readability.
[Introduction] The comparison with Genzel-Stollenwerk (2023) is stated in the abstract; a short paragraph contrasting the new uniform rates with their nonuniform ones would help situate the improvement.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the positive assessment, the recommendation for minor revision, and the constructive comments. We address the major comments point by point below and will incorporate the requested clarifications into the revised manuscript.

read point-by-point responses

Referee: [main recovery theorem and covering argument for uniform RAIC] The central claim that the extra uniformity terms remain negligible (and do not change the order of the recovery error) in the three canonical settings rests on the explicit dependence of the RAIC constants on the covering number of the structured class; this dependence must be tracked through the PGD convergence analysis to confirm negligibility.

Authors: We appreciate the referee's emphasis on explicit tracking. The manuscript establishes uniform RAIC via covering arguments and shows that, in the three settings of IHT, ℓ₁-projection for sparse signals, and ℓ₁-projection for approximately sparse signals, the resulting additive terms are absorbed into logarithmic factors relative to the nonuniform baselines. To address the request directly, the revised version will augment the proof of the main recovery theorem with a step-by-step propagation of the covering-number dependence through the PGD error recursion, confirming that it contributes only the stated logarithmic overhead and does not alter the order of the recovery error. revision: yes
Referee: [robustness extension paragraph] The robustness statement is presented as following directly from a single additional random-process bound; the precise assumptions on the noise and corruption (e.g., sub-Gaussian tails, bounded support) that make this bound uniform over the signal class should be stated explicitly so that the extension does not introduce hidden structural requirements.

Authors: We agree that the assumptions should be stated explicitly for clarity. The robustness result follows from controlling one additional gradient-mismatch random process uniformly over the signal class. In the revised manuscript we will add an explicit statement of the assumptions (sub-Gaussian noise with appropriate variance proxy and bounded corruption) in the robustness theorem and the surrounding paragraph, ensuring the uniformity holds under these standard conditions without introducing further structural requirements on the signal class. revision: yes

Circularity Check

0 steps flagged

Derivation self-contained; no circular reductions identified

full rationale

The central chain defines the restricted approximate invertibility condition (RAIC) on the gradient of the scaled ℓ2 loss, then verifies the uniform RAIC over the structured class via covering (or VC-dimension) arguments under piecewise Lipschitz link functions. Uniform recovery then follows from PGD convergence under this condition. The argument explicitly builds on external nonuniform recovery results (Plan-Vershynin 2016, Oymak-Soltanolkotabi 2017) and standard concentration/covering tools; the extra uniformity error terms are bounded separately and shown negligible in the three canonical projection settings. No step reduces by construction to a fitted parameter, self-defined quantity, or load-bearing self-citation chain. The robustness extension is a direct additional-process bound without new structural assumptions.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on the existence of uniform RAIC for the gradient under the stated link-function class; this is proved rather than postulated, but the proof technique (covering number bounds) is standard and not invented here.

axioms (2)

domain assumption Gaussian measurements and piecewise Lipschitz link functions
Invoked to establish uniform RAIC via covering argument for the single-index model.
standard math Standard concentration inequalities for random processes
Used implicitly to control the gradient mismatch and noise terms.

pith-pipeline@v0.9.0 · 5643 in / 1465 out tokens · 37353 ms · 2026-05-09T23:47:14.109966+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

300 extracted references · 14 canonical work pages

[1]

Foundations of Computational Mathematics , volume=

Sharp recovery bounds for convex demixing, with applications , author=. Foundations of Computational Mathematics , volume=. 2014 , publisher=

2014
[2]

Applied and Computational Harmonic Analysis , volume=

An algebraic characterization of injectivity in phase retrieval , author=. Applied and Computational Harmonic Analysis , volume=. 2015 , publisher=

2015
[3]

Proceedings of Thirty Eighth Conference on Learning Theory , pages =

Learning sparse generalized linear models with binary outcomes via iterative hard thresholding , author =. Proceedings of Thirty Eighth Conference on Learning Theory , pages =. 2025 , volume =

2025
[4]

Sampling theory, a renaissance , pages=

Convex recovery of a structured signal from independent random linear measurements , author=. Sampling theory, a renaissance , pages=. 2015 , publisher=

2015
[5]

IEEE Transactions on Signal Processing , volume=

On unlimited sampling and reconstruction , author=. IEEE Transactions on Signal Processing , volume=. 2020 , publisher=

2020
[6]

In Preparation , year=

Algorithms for Matrix Phase Retrieval , author=. In Preparation , year=
[7]

Applied and Computational Harmonic Analysis , volume=

Generalized phase retrieval: measurement number, matrix recovery and beyond , author=. Applied and Computational Harmonic Analysis , volume=. 2019 , publisher=

2019
[8]

Applied and Computational Harmonic Analysis , volume=

Phase retrieval for sparse signals , author=. Applied and Computational Harmonic Analysis , volume=. 2014 , publisher=

2014
[9]

Applied and Computational Harmonic Analysis , volume=

CoSaMP: Iterative signal recovery from incomplete and inaccurate samples , author=. Applied and Computational Harmonic Analysis , volume=. 2009 , publisher=

2009
[10]

IEEE Transactions on Signal Processing , volume=

Beyond Procrustes: Balancing-free gradient descent for asymmetric low-rank matrix sensing , author=. IEEE Transactions on Signal Processing , volume=. 2021 , publisher=

2021
[11]

SIAM Journal on Matrix Analysis and Applications , volume=

Dynamical approximation by hierarchical Tucker and tensor-train tensors , author=. SIAM Journal on Matrix Analysis and Applications , volume=. 2013 , publisher=

2013
[12]

Technical Report , year=

The singular value decomposition and low rank approximation , author=. Technical Report , year=
[13]

Fast low-rank estimation by projected gradient descent: General statistical and algorithmic guarantees.arXiv preprint arXiv:1509.03025, 2015

Fast low-rank estimation by projected gradient descent: General statistical and algorithmic guarantees , author=. arXiv preprint arXiv:1509.03025 , year=

work page arXiv
[14]

SIAM Review , volume=

Tensor decompositions and applications , author=. SIAM Review , volume=. 2009 , publisher=

2009
[15]

IEEE Transactions on Information Theory , volume=

Sample-efficient low rank phase retrieval , author=. IEEE Transactions on Information Theory , volume=. 2021 , publisher=

2021
[16]

Advances in Neural Information Processing Systems , volume=

Fast algorithms for robust PCA via gradient descent , author=. Advances in Neural Information Processing Systems , volume=
[17]

International Conference on Machine Learning , year=

Guarantees of a preconditioned subgradient algorithm for overparameterized asymmetric low-rank matrix recovery , author=. International Conference on Machine Learning , year=
[18]

International Conference on Machine Learning , pages=

No spurious local minima in nonconvex low rank problems: A unified geometric analysis , author=. International Conference on Machine Learning , pages=. 2017 , organization=

2017
[19]

Advances in Neural Information Processing Systems , volume=

A convergent gradient descent algorithm for rank minimization and semidefinite programming from random linear measurements , author=. Advances in Neural Information Processing Systems , volume=
[20]

The Annals of Statistics , volume=

An optimal statistical and computational framework for generalized tensor estimation , author=. The Annals of Statistics , volume=. 2022 , publisher=

2022
[21]

Information and Inference: A Journal of the IMA , volume=

Phase retrieval of low-rank matrices by anchored regression , author=. Information and Inference: A Journal of the IMA , volume=. 2021 , publisher=

2021
[22]

IEEE Transactions on Signal Processing , volume=

Low-rank phase retrieval , author=. IEEE Transactions on Signal Processing , volume=. 2017 , publisher=

2017
[23]

SIAM Journal on Matrix Analysis and Applications , volume=

A multilinear singular value decomposition , author=. SIAM Journal on Matrix Analysis and Applications , volume=. 2000 , publisher=

2000
[24]

The Annals of Statistics , volume=

Tensor-on-tensor regression: Riemannian optimization, over-parameterization, statistical-computational gap and their interplay , author=. The Annals of Statistics , volume=. 2024 , publisher=

2024
[25]

SIAM Journal on Mathematics of Data Science , volume=

ISLET: Fast and optimal low-rank tensor regression via importance sketching , author=. SIAM Journal on Mathematics of Data Science , volume=. 2020 , publisher=

2020
[26]

Mathematical Programming , volume=

Misspecified nonconvex statistical optimization for sparse phase retrieval , author=. Mathematical Programming , volume=. 2019 , publisher=

2019
[27]

arXiv preprint arXiv:2407.04951 , year=

Optimal quantized compressed sensing via projected gradient descent , author=. arXiv preprint arXiv:2407.04951 , year=

work page arXiv
[28]

Advances in Neural Information Processing Systems , volume=

Agnostic estimation for misspecified phase retrieval models , author=. Advances in Neural Information Processing Systems , volume=
[29]

IEEE International Symposium on Information Theory , year=

Phase Transitions in Phase-Only Compressed Sensing , author=. IEEE International Symposium on Information Theory , year=
[30]

Proceedings of the IEEE , volume=

The importance of phase in signals , author=. Proceedings of the IEEE , volume=. 1981 , publisher=

1981
[31]

SIAM Journal on Applied Mathematics , volume=

Signal reconstruction from phase-only measurements: Uniqueness condition, minimal measurement number and beyond , author=. SIAM Journal on Applied Mathematics , volume=. 2023 , publisher=

2023
[32]

Optik , volume=

A practical algorithm for the determination of phase from image and diffraction plane pictures , author=. Optik , volume=
[33]

Doklady Akademii Nauk , volume=

Gaussian random processes and measures of solid angles in Hilbert space , author=. Doklady Akademii Nauk , volume=. 1971 , organization=

1971
[34]

IEEE Transactions on Information Theory , volume=

Compressed sensing with nonlinear observations and related nonlinear optimization problems , author=. IEEE Transactions on Information Theory , volume=. 2013 , publisher=

2013
[35]

Foundations of Computational Mathematics , volume=

Solving quadratic equations via PhaseLift when there are about as many equations as unknowns , author=. Foundations of Computational Mathematics , volume=. 2014 , publisher=

2014
[36]

Applied and Computational Harmonic Analysis , volume=

Saving phase: Injectivity and stability for phase retrieval , author=. Applied and Computational Harmonic Analysis , volume=. 2014 , publisher=

2014
[37]

Journal of the American Statistical Association , volume=

Generalized low-rank plus sparse tensor estimation by fast Riemannian optimization , author=. Journal of the American Statistical Association , volume=. 2023 , publisher=

2023
[38]

The Annals of Probability , pages=

Decoupling inequalities for the tail probabilities of multivariate U-statistics , author=. The Annals of Probability , pages=. 1995 , publisher=

1995
[39]

arXiv preprint arXiv:2007.08904 , year=

Provable near-optimal low-multilinear-rank tensor recovery , author=. arXiv preprint arXiv:2007.08904 , year=

work page arXiv 2007
[40]

SIAM Journal on Matrix Analysis and Applications , volume=

Dynamical tensor approximation , author=. SIAM Journal on Matrix Analysis and Applications , volume=. 2010 , publisher=

2010
[41]

Proceedings of the IEEE , volume=

Extension of PCA to higher order data structures: An introduction to tensors, tensor decompositions, and tensor PCA , author=. Proceedings of the IEEE , volume=. 2018 , publisher=

2018
[42]

2018 , publisher=

An introduction to generalized linear models , author=. 2018 , publisher=

2018
[43]

Proceedings of the National Academy of Sciences , volume=

A modern maximum-likelihood theory for high-dimensional logistic regression , author=. Proceedings of the National Academy of Sciences , volume=. 2019 , publisher=

2019
[44]

IEEE Transactions on Information Theory , volume=

Model selection and minimax estimation in generalized linear models , author=. IEEE Transactions on Information Theory , volume=. 2016 , publisher=

2016
[45]

2009 , publisher=

Tensors in image processing and computer vision , author=. 2009 , publisher=

2009
[46]

IEEE Transactions on Signal Processing , volume=

Tensor decomposition for signal processing and machine learning , author=. IEEE Transactions on Signal Processing , volume=. 2017 , publisher=

2017
[47]

Journal of Visual Communication and Image Representation , volume=

Decomposition-based tensor learning regression for improved classification of multimedia , author=. Journal of Visual Communication and Image Representation , volume=. 2016 , publisher=

2016
[48]

Bayesian Analysis , volume=

Variational Bayesian Logistic Tensor Regression with Application to Image Recognition , author=. Bayesian Analysis , volume=. 2025 , publisher=

2025
[49]

International Conference on Intelligent Science and Intelligent Data Engineering , pages=

Logistic tensor regression for classification , author=. International Conference on Intelligent Science and Intelligent Data Engineering , pages=. 2012 , organization=

2012
[50]

Statistical Methods in Medical Research , volume=

Bayesian tensor logistic regression with applications to neuroimaging data analysis of Alzheimer’s disease , author=. Statistical Methods in Medical Research , volume=. 2022 , publisher=

2022
[51]

Rigollet and J.-C

High-dimensional statistics , author=. arXiv preprint arXiv:2310.19244 , year=

work page arXiv
[52]

2006 , Booktitle =

Elements of Information Theory , Author =. 2006 , Booktitle =

2006
[53]

Advances in Neural Information Processing Systems , volume=

Minimax bounds for generalized linear models , author=. Advances in Neural Information Processing Systems , volume=
[54]

Transactions on Machine Learning Research , year=

Structured low-rank tensors for generalized linear models , author=. Transactions on Machine Learning Research , year=
[55]

International Conference on Artificial Intelligence and Statistics , pages=

Sparse and low-rank tensor estimation via cubic sketchings , author=. International Conference on Artificial Intelligence and Statistics , pages=. 2020 , organization=

2020
[56]

IEEE Transactions on Information Theory , volume=

Tensor SVD: Statistical and computational limits , author=. IEEE Transactions on Information Theory , volume=. 2018 , publisher=

2018
[57]

Communications on Pure and Applied Mathematics , volume=

Spectral algorithms for tensor completion , author=. Communications on Pure and Applied Mathematics , volume=. 2018 , publisher=

2018
[58]

Journal of Machine Learning Research , volume=

Bayesian tensor regression , author=. Journal of Machine Learning Research , volume=
[59]

The Annals of Applied Statistics , volume=

Multilinear tensor regression for longitudinal relational data , author=. The Annals of Applied Statistics , volume=
[60]

2010 IEEE International Conference on Image Processing , pages=

Tensor completion for on-board compression of hyperspectral images , author=. 2010 IEEE International Conference on Image Processing , pages=. 2010 , organization=

2010
[61]

Journal of the American Statistical Association , volume=

Tensor regression with applications in neuroimaging data analysis , author=. Journal of the American Statistical Association , volume=. 2013 , publisher=

2013
[62]

International Conference on Learning Representations , year=

Generative Principal Component Analysis , author=. International Conference on Learning Representations , year=
[63]

Sankhya A , pages=

Finite sample rates for logistic regression with small noise or few samples , author=. Sankhya A , pages=. 2024 , publisher=

2024
[64]

Subspace estimation from unbalanced and incomplete data matrices: _

Cai, Changxiao and Li, Gen and Chi, Yuejie and Poor, H Vincent and Chen, Yuxin , journal=. Subspace estimation from unbalanced and incomplete data matrices: _
[65]

2009 , publisher=

The elements of statistical learning , author=. 2009 , publisher=

2009
[66]

Annals of Statistics , year=

Multilayer tensor factorization with applications to recommender systems , author=. Annals of Statistics , year=
[67]

2008 , publisher=

Applied multiway data analysis , author=. 2008 , publisher=

2008
[68]

Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=

Regression shrinkage and selection via the lasso , author=. Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=. 1996 , publisher=

1996
[69]

IEEE Transactions on Image Processing , volume=

Tensor-based formulation and nuclear norm regularization for multienergy computed tomography , author=. IEEE Transactions on Image Processing , volume=. 2014 , publisher=

2014
[70]

Advances in Neural Information Processing Systems , volume=

Tensor decomposition for fast parsing with latent-variable PCFGs , author=. Advances in Neural Information Processing Systems , volume=
[71]

IEEE Transactions on Audio, Speech, and Language Processing , volume=

Discrimination of speech from nonspeech based on multiscale spectro-temporal modulations , author=. IEEE Transactions on Audio, Speech, and Language Processing , volume=. 2006 , publisher=

2006
[72]

Ongoing Work , year=

An instance optimal algorithm for sparse phase retrieval , author=. Ongoing Work , year=
[73]

Indian Pines Hyperspectral Scene , howpublished =
[74]

Foundations of Computational Mathematics , volume=

Low-rank matrix recovery with composite optimization: good conditioning and rapid convergence , author=. Foundations of Computational Mathematics , volume=. 2021 , publisher=

2021
[75]

2019 , publisher=

High-dimensional statistics: A non-asymptotic viewpoint , author=. 2019 , publisher=

2019
[76]

Information and Inference: A Journal of the IMA , volume=

Alternating minimization for generalized rank-1 matrix sensing: sharp predictions from a random initialization , author=. Information and Inference: A Journal of the IMA , volume=. 2024 , publisher=

2024
[77]

, author=

Tensor decompositions for learning latent variable models. , author=. J. Mach. Learn. Res. , volume=
[78]

Journal of the American Statistical Association , volume=

Variable selection via nonconcave penalized likelihood and its oracle properties , author=. Journal of the American Statistical Association , volume=. 2001 , publisher=

2001
[79]

2011 , publisher=

Statistics for high-dimensional data: methods, theory and applications , author=. 2011 , publisher=

2011
[80]

Annual review of statistics and its application , volume=

Tensors in statistics , author=. Annual review of statistics and its application , volume=. 2021 , publisher=

2021

Showing first 80 references.