pith. sign in

arxiv: 2605.19364 · v1 · pith:7U643OWFnew · submitted 2026-05-19 · 🧮 math.ST · math.PR· stat.TH

Optimal Spectral Algorithms for Correlated Two-view Models in High Dimensions

Hang Du , Henry Hu , Saba Lepsveridze This is my paper

Pith reviewed 2026-05-20 02:33 UTC · model grok-4.3

classification 🧮 math.ST math.PRstat.TH
keywords spectral algorithmshigh-dimensional inferencecorrelated modelsstrong detectionweak recoverycanonical correlation analysisspiked Wigner modelWishart model
0
0 comments X

The pith

Spectral algorithms achieve strong detection and weak recovery down to information-theoretic thresholds in correlated two-view models without knowing the parameters.

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

The paper constructs explicit spectral algorithms for three high-dimensional correlated two-view models that attain the optimal thresholds for strong detection and weak recovery. It proves matching information-theoretic lower bounds and shows the algorithms require only the observed data. A sympathetic reader would care because the work unifies canonical correlation analysis with spiked Wigner and Wishart models under one framework and demonstrates that simple spectral procedures can be optimal. The approach relies on a TAP-type heuristic to motivate the constructions. This establishes the broad applicability of spectral methods in these settings.

Core claim

In high-dimensional canonical correlation analysis and the correlated spiked Wigner and Wishart models, explicit spectral algorithms achieve strong detection and weak recovery down to the corresponding thresholds, with matching information-theoretic lower bounds, and operate without knowledge of the model parameters by relying solely on the observed data.

What carries the argument

A general framework motivated by the TAP-type heuristic from statistical physics that unifies the three models and guides construction of the parameter-free spectral algorithms.

If this is right

  • Spectral methods are optimal for strong detection and weak recovery in these three models.
  • The algorithms apply directly to high-dimensional canonical correlation analysis without parameter tuning.
  • Matching lower bounds confirm that the thresholds cannot be improved by any method.
  • The procedures extend to both Wigner and Wishart type correlated models under the same framework.

Where Pith is reading between the lines

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

  • The same framework could guide algorithm design for other multi-view or multi-modal high-dimensional inference problems.
  • Practical implementations might test the parameter-free property on real correlated datasets from genomics or finance.
  • The optimality result suggests spectral methods deserve priority in related statistical physics motivated models.

Load-bearing premise

The TAP-type heuristic from statistical physics provides a valid and tight characterization of the detection and recovery thresholds achievable by spectral methods.

What would settle it

A direct calculation or simulation on the correlated spiked Wigner model showing that the proposed spectral algorithm fails to achieve detection or weak recovery exactly at the threshold value predicted by the TAP heuristic would disprove the optimality claim.

Figures

Figures reproduced from arXiv: 2605.19364 by Hang Du, Henry Hu, Saba Lepsveridze.

Figure 1
Figure 1. Figure 1: The panels show numerical simulations for the CCA model with parameters n = 10000, m = k = 5000, and threshold κ = 0.707. The left panel shows the spectrum of the CCA estimator W with planted correlation ρ = 0.85. The right panel shows the overlap, |⟨w, ˆ (a, b)⟩|2 , averaged over 50 trials, as a function of ρ. In [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The two panels plot numerical simulations of the overlap, |⟨w, ˆ (a, b)⟩|2 , as a function of ρ. The left panel shows simulations for CSWig with parameters n = 5000, signal strengths α = β = 0.8, and threshold κ = 0.75, averaged over 12 trials. The right panel shows simulations for CSWish with parameters n = 10000, m = 5000, signal strengths α = β = 0.6, and threshold κ = 0.624, averaged over 24 trials. In… view at source ↗
read the original abstract

We study high-dimensional inference in correlated two-view models, focusing on spectral methods for strong detection and weak recovery. We introduce a general framework, motivated by a TAP type heuristic from statistical physics, that provides a unified treatment of three canonical models: high-dimensional canonical correlation analysis, and the correlated spiked Wigner and Wishart models. Our main contribution is to construct explicit spectral algorithms in all three settings, that achieve strong detection and weak recovery down to the corresponding thresholds, where we prove matching information-theoretic lower bounds. Furthermore, our spectral procedures operate without knowledge of the model parameters, relying solely on the observed data. This demonstrates the optimality of spectral methods in these models and the broad statistical applicability of the framework.

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.

Referee Report

1 major / 3 minor

Summary. The paper introduces a general framework, motivated by a TAP-type heuristic, for spectral methods in high-dimensional correlated two-view models. It constructs explicit spectral algorithms for high-dimensional CCA, the correlated spiked Wigner model, and the correlated spiked Wishart model. These algorithms achieve strong detection and weak recovery down to the information-theoretic thresholds, with matching lower bounds proved via second-moment methods and contiguity arguments. The procedures are parameter-free and rely solely on the observed data matrices.

Significance. If the results hold, this work is significant for establishing the optimality of simple, data-driven spectral methods in three canonical correlated inference problems. The explicit constructions, parameter-free operation, and tight matching of upper and lower bounds provide concrete evidence that spectral algorithms attain the information-theoretic limits without external parameter inputs, strengthening the case for their broad applicability in high-dimensional statistics.

major comments (1)
  1. [§4.2] §4.2, Theorem 4.3: the contiguity argument for the lower bound on weak recovery in the correlated spiked Wishart model uses a specific second-moment calculation; it is unclear whether the variance of the test statistic remains bounded away from zero exactly at the threshold when the correlation parameter approaches its critical value from above.
minor comments (3)
  1. [§2.3] §2.3: the notation for the two-view covariance matrices is introduced separately for each model; a unified table comparing the observation models across CCA, Wigner, and Wishart would improve readability.
  2. [Figure 2] Figure 2: the empirical phase-transition curves are shown without overlaid theoretical thresholds; adding dashed lines at the predicted critical values would make the matching between theory and simulation more immediate.
  3. [§5.1] §5.1: a few instances of 'parameter-free' appear without an explicit forward reference to the data-driven normalization step; a single clarifying sentence would prevent minor confusion.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We are grateful to the referee for the positive assessment and the detailed comment on our manuscript. We address the major comment below and plan to incorporate clarifications in the revised version.

read point-by-point responses

  1. Referee: [§4.2] §4.2, Theorem 4.3: the contiguity argument for the lower bound on weak recovery in the correlated spiked Wishart model uses a specific second-moment calculation; it is unclear whether the variance of the test statistic remains bounded away from zero exactly at the threshold when the correlation parameter approaches its critical value from above.

    Authors: We thank the referee for pointing this out. In the proof of Theorem 4.3, the second-moment method is used to establish contiguity between the null and alternative measures. The calculation of the second moment involves the correlation parameter ρ, and we verify that as ρ approaches the critical value ρ_c from above, the variance of the relevant test statistic (derived from the log-likelihood ratio or the second-moment quantity) remains bounded by a constant independent of the dimension and other parameters. Specifically, the expression for the variance is shown to be O(1) uniformly in a neighborhood of the threshold. To make this explicit, we will add a short remark or lemma in the revised manuscript detailing the uniform bound on the variance as ρ ↓ ρ_c. This ensures the contiguity holds at the threshold. revision: yes

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper's central derivation constructs explicit spectral algorithms for CCA, correlated spiked Wigner, and correlated spiked Wishart models that achieve strong detection and weak recovery at information-theoretic thresholds, with matching lower bounds proved directly via second-moment methods and contiguity arguments. These proofs and the data-driven, parameter-free nature of the algorithms (using only observed matrices to form quadratic forms or eigenvectors) are independent of the TAP heuristic, which appears solely in the motivation section for candidate algorithm derivation. No self-definitional reductions, fitted inputs renamed as predictions, load-bearing self-citations, or ansatz smuggling occur in the load-bearing steps; the framework remains self-contained with external falsifiability through the proved bounds.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Based on the abstract, the central claims rest on the validity of the TAP-type heuristic for characterizing thresholds in these models and on standard high-dimensional asymptotic assumptions for the three canonical models; no explicit free parameters or invented entities are mentioned.

axioms (2)
  • domain assumption TAP-type heuristic from statistical physics accurately predicts the information-theoretic thresholds for strong detection and weak recovery in correlated two-view models
    Abstract states the framework is motivated by this heuristic and uses it to construct algorithms that achieve the thresholds.
  • domain assumption The three models (high-dimensional CCA, correlated spiked Wigner, correlated spiked Wishart) share a common structure amenable to unified spectral treatment in the high-dimensional regime
    Abstract presents a general framework that provides unified treatment of these three canonical models.

pith-pipeline@v0.9.0 · 5646 in / 1528 out tokens · 35951 ms · 2026-05-20T02:33:27.283024+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

45 extracted references · 45 canonical work pages · 2 internal anchors

  1. [1]

    Singularities of solutions to qua- dratic vector equations on the complex upper half-plane

    Oskari Ajanki, L´ aszl´ o Erd˝ os, and Torben Kr¨ uger. “Singularities of solutions to qua- dratic vector equations on the complex upper half-plane”. In:Comm. Pure Appl. Math. 70.9 (2017), pp. 1672–1705.issn: 0010-3640,1097-0312.doi:10.1002/cpa.21639.url: https://doi.org/10.1002/cpa.21639

    work page doi:10.1002/cpa.21639.url: 2017

  2. [2]

    Stability of the matrix Dyson equation and random matrices with correlations

    Oskari H. Ajanki, L´ aszl´ o Erd˝ os, and Torben Kr¨ uger. “Stability of the matrix Dyson equation and random matrices with correlations”. In:Probab. Theory Related Fields 173.1-2 (2019), pp. 293–373.issn: 0178-8051,1432-2064.doi:10.1007/s00440-018- 0835-z.url:https://doi.org/10.1007/s00440-018-0835-z

    work page doi:10.1007/s00440-018- 2019

  3. [3]

    No eigenvalues outside the support of the limiting spectral distribution of large-dimensional sample covariance matrices

    Z. D. Bai and Jack W. Silverstein. “No eigenvalues outside the support of the limiting spectral distribution of large-dimensional sample covariance matrices”. In:Ann. Probab. 26.1 (1998), pp. 316–345.issn: 0091-1798,2168-894X.doi:10.1214/aop/1022855421. url:https://doi.org/10.1214/aop/1022855421

    work page doi:10.1214/aop/1022855421 1998

  4. [5]

    In: International Conference on Social Robotics

    Zhidong Bai and Jack W. Silverstein.Spectral analysis of large dimensional random matrices. Second. Springer Series in Statistics. Springer, New York, 2010, pp. xvi+551. isbn: 978-1-4419-0660-1.doi:10.1007/978- 1- 4419- 0661- 8.url:https://doi. org/10.1007/978-1-4419-0661-8

    work page doi:10.1007/978- 2010

  5. [6]

    Phase transition of the largest eigenvalue for nonnull complex sample covariance matrices

    Jinho Baik, G´ erard Ben Arous, and Sandrine P´ ech´ e. “Phase transition of the largest eigenvalue for nonnull complex sample covariance matrices”. In:Ann. Probab.33.5 (2005), pp. 1643–1697.issn: 0091-1798,2168-894X.doi:10.1214/009117905000000233. url:https://doi.org/10.1214/009117905000000233

    work page doi:10.1214/009117905000000233 2005

  6. [7]

    The eigenvalues and eigenvectors of finite, low rank perturbations of large random matrices

    Florent Benaych-Georges and Raj Rao Nadakuditi. “The eigenvalues and eigenvectors of finite, low rank perturbations of large random matrices”. In:Adv. Math.227.1 (2011), pp. 494–521.issn: 0001-8708,1090-2082.doi:10.1016/j.aim.2011.02.007.url: https://doi.org/10.1016/j.aim.2011.02.007. 134 REFERENCES

    work page doi:10.1016/j.aim.2011.02.007.url: 2011

  7. [8]

    Reducibility and Computational Lower Bounds for Problems with Planted Sparse Structure

    Matthew Brennan, Guy Bresler, and Wasim Huleihel. “Reducibility and Computational Lower Bounds for Problems with Planted Sparse Structure”. In:Proceedings of the 31st Conference On Learning Theory. Ed. by S´ ebastien Bubeck, Vianney Perchet, and Philippe Rigollet. Vol. 75. Proceedings of Machine Learning Research. PMLR, 2018, pp. 48–166.url:https://procee...

    work page 2018

  8. [9]

    Anna Bykhovskaya and Vadim Gorin.High-Dimensional Canonical Correlation Analy- sis. 2025. arXiv:2306.16393 [econ.EM].url:https://arxiv.org/abs/2306.16393

    work page arXiv 2025

  9. [10]

    Sparse PCA: optimal rates and adaptive estimation

    T. Tony Cai, Zongming Ma, and Yihong Wu. “Sparse PCA: optimal rates and adaptive estimation”. In:Ann. Statist.41.6 (2013), pp. 3074–3110.issn: 0090-5364,2168-8966. doi:10.1214/13-AOS1178.url:https://doi.org/10.1214/13-AOS1178

    work page doi:10.1214/13-aos1178.url:https://doi.org/10.1214/13-aos1178 2013

  10. [11]

    The largest eigenvalue of rank one deformation of large Wigner matrices

    Delphine F´ eral and Sandrine P´ ech´ e. “The largest eigenvalue of rank one deformation of large Wigner matrices”. In:Comm. Math. Phys.272.1 (2007), pp. 185–228.issn: 0010-3616,1432-0916.doi:10.1007/s00220- 007- 0209- 3.url:https://doi.org/ 10.1007/s00220-007-0209-3

    work page doi:10.1007/s00220- 2007

  11. [12]

    Necessary and sufficient condition that the limit of Stieltjes transforms is a Stieltjes transform

    Jeffrey S. Geronimo and Theodore P. Hill. “Necessary and sufficient condition that the limit of Stieltjes transforms is a Stieltjes transform”. In:J. Approx. Theory121.1 (2003), pp. 54–60.issn: 0021-9045,1096-0430.doi:10.1016/S0021-9045(02)00042-4. url:https://doi.org/10.1016/S0021-9045(02)00042-4

    work page doi:10.1016/s0021-9045(02)00042-4 2003

  12. [13]

    Some inequalities for Gaussian processes and applications

    Yehoram Gordon. “Some inequalities for Gaussian processes and applications”. In: Israel Journal of Mathematics50.4 (1985), pp. 265–289.issn: 1565-8511.doi:10 . 1007/BF02759761.url:https://doi.org/10.1007/BF02759761

    work page doi:10.1007/bf02759761 1985

  13. [14]

    On bilinear forms based on the resolvent of large random matri- ces

    Walid Hachem et al. “On bilinear forms based on the resolvent of large random matri- ces”. In:Ann. Inst. Henri Poincar´ e Probab. Stat.49.1 (2013), pp. 36–63.issn: 0246- 0203,1778-7017.doi:10.1214/11- AIHP450.url:https://doi.org/10.1214/11- AIHP450

    work page doi:10.1214/11- 2013

  14. [15]

    On the distribution of the largest eigenvalue in principal compo- nents analysis

    Iain M. Johnstone. “On the distribution of the largest eigenvalue in principal compo- nents analysis”. In:Ann. Statist.29.2 (2001), pp. 295–327.issn: 0090-5364,2168-8966. doi:10.1214/aos/1009210544.url:https://doi.org/10.1214/aos/1009210544

    work page doi:10.1214/aos/1009210544.url:https://doi.org/10.1214/aos/1009210544 2001

  15. [16]

    Optimal thresholds and algorithms for a model of multi-modal learning in high dimensions

    Christian Keup and Lenka Zdeborov´ a. “Optimal thresholds and algorithms for a model of multi-modal learning in high dimensions”. In:J. Stat. Mech. Theory Exp.9 (2025), Paper No. 093302, 37.issn: 1742-5468

    work page 2025

  16. [17]

    MMSE of probabilistic low- rank matrix estimation: Universality with respect to the output channel

    Thibault Lesieur, Florent Krzakala, and Lenka Zdeborova. “MMSE of probabilistic low- rank matrix estimation: Universality with respect to the output channel”. In:2015 53rd Annual Allerton Conference on Communication, Control, and Computing (Allerton). IEEE, 2015, pp. 680–687.doi:10.1109/allerton.2015.7447070.url:http://dx. doi.org/10.1109/ALLERTON.2015.7447070

    work page doi:10.1109/allerton.2015.7447070.url:http://dx 2015

  17. [18]

    Phase transitions in sparse PCA

    Thibault Lesieur, Florent Krzakala, and Lenka Zdeborova. “Phase transitions in sparse PCA”. In:2015 IEEE International Symposium on Information Theory (ISIT). IEEE, 2015, pp. 1635–1639.doi:10.1109/isit.2015.7282733.url:http://dx.doi.org/ 10.1109/ISIT.2015.7282733

    work page doi:10.1109/isit.2015.7282733.url:http://dx.doi.org/ 2015

  18. [19]

    Constrained low-rank ma- trix estimation: phase transitions, approximate message passing and applications

    Thibault Lesieur, Florent Krzakala, and Lenka Zdeborov´ a. “Constrained low-rank ma- trix estimation: phase transitions, approximate message passing and applications”. In: Journal of Statistical Mechanics: Theory and Experiment2017.7 (2017), p. 073403. issn: 1742-5468.doi:10.1088/1742-5468/aa7284.url:http://dx.doi.org/10. 1088/1742-5468/aa7284. REFERENCES 135

    work page doi:10.1088/1742-5468/aa7284.url:http://dx.doi.org/10 2017

  19. [20]

    Statistical and Computational Phase Transitions in Spiked Tensor Estimation

    Thibault Lesieur et al. “Statistical and Computational Phase Transitions in Spiked Tensor Estimation”. In:Proceedings of the IEEE International Symposium on Infor- mation Theory (ISIT). IEEE, 2017, pp. 511–515

    work page 2017

  20. [21]

    Zhangsong Li.A Smooth Computational Transition in Tensor PCA. 2025. arXiv:2509. 09904 [math.ST].url:https://arxiv.org/abs/2509.09904

    work page arXiv 2025

  21. [22]

    Zhangsong Li.The Algorithmic Phase Transition in Correlated Spiked Models. 2025. arXiv:2511.06040 [math.ST].url:https://arxiv.org/abs/2511.06040

    work page arXiv 2025

  22. [23]

    Joint and individual variation explained (JIVE) for integrated analysis of multiple data types

    Eric F. Lock et al. “Joint and individual variation explained (JIVE) for integrated analysis of multiple data types”. In:Ann. Appl. Stat.7.1 (2013), pp. 523–542.issn: 1932-6157,1941-7330.doi:10.1214/12-AOAS597.url:https://doi.org/10.1214/ 12-AOAS597

    work page doi:10.1214/12-aoas597.url:https://doi.org/10.1214/ 2013

  23. [24]

    Preprint, arXiv:2510.17561

    Pierre Mergny and Lenka Zdeborov´ a.Spectral Thresholds in Correlated Spiked Models and Fundamental Limits of Partial Least Squares. Preprint, arXiv:2510.17561. 2025

    work page arXiv 2025

  24. [25]

    Mean-field message-passing equations in the Hopfield model and its generalizations

    Marc M´ ezard. “Mean-field message-passing equations in the Hopfield model and its generalizations”. In:Phys. Rev. E95.2 (2017), pp. 022117, 15.issn: 2470-0045,2470- 0053.doi:10 . 1103 / physreve . 95 . 022117.url:https : / / doi . org / 10 . 1103 / physreve.95.022117

    work page 2017

  25. [26]

    Algebraic

    Marc M´ ezard and Andrea Montanari.Information, physics, and computation. Oxford Graduate Texts. Oxford University Press, Oxford, 2009, pp. xiv+569.isbn: 978-0- 19-857083-7.doi:10 . 1093 / acprof : oso / 9780198570837 . 001 . 0001.url:https : //doi.org/10.1093/acprof:oso/9780198570837.001.0001

    work page doi:10.1093/acprof:oso/9780198570837.001.0001 2009

  26. [27]

    A Friendly Tutorial on Mean-Field Spin Glass Techniques for Non-Physicists

    Andrea Montanari and Subhabrata Sen. “A Friendly Tutorial on Mean-Field Spin Glass Techniques for Non-Physicists”. In:Foundations and Trends in Machine Learning17.1 (2024), pp. 1–173.doi:10.1561/2200000105

    work page doi:10.1561/2200000105 2024

  27. [28]

    Reconstruction and estimation in the planted partition model

    Elchanan Mossel, Joe Neeman, and Allan Sly. “Reconstruction and estimation in the planted partition model”. In:Probab. Theory Related Fields162.3-4 (2015), pp. 431– 461.issn: 0178-8051,1432-2064.doi:10 . 1007 / s00440 - 014 - 0576 - 6.url:https : //doi.org/10.1007/s00440-014-0576-6

    work page doi:10.1007/s00440-014-0576-6 2015

  28. [29]

    Optimality and sub-optimality of PCA I: Spiked random matrix models

    Amelia Perry et al. “Optimality and sub-optimality of PCA I: Spiked random matrix models”. In:Ann. Statist.46.5 (2018), pp. 2416–2451.issn: 0090-5364,2168-8966.doi: 10.1214/17-AOS1625.url:https://doi.org/10.1214/17-AOS1625

    work page doi:10.1214/17-aos1625.url:https://doi.org/10.1214/17-aos1625 2018

  29. [30]

    Cam- bridge University Press, 2025

    Yury Polyanskiy and Yihong Wu.Information Theory: From Coding to Learning. Cam- bridge University Press, 2025

    work page 2025

  30. [31]

    A Statistical Model for Tensor PCA

    Emile Richard and Andrea Montanari. “A Statistical Model for Tensor PCA”. In:Ad- vances in Neural Information Processing Systems. Vol. 27. MIT Press, 2014, pp. 2897– 2905

    work page 2014

  31. [32]

    Hanson-Wright inequality and sub-Gaussian concentration

    Mark Rudelson and Roman Vershynin. “Hanson-Wright inequality and sub-Gaussian concentration”. In:Electron. Commun. Probab.18 (2013), no. 82, 9.issn: 1083-589X. doi:10.1214/ECP.v18-2865.url:https://doi.org/10.1214/ECP.v18-2865

    work page doi:10.1214/ecp.v18-2865.url:https://doi.org/10.1214/ecp.v18-2865 2013

  32. [33]

    Computational barriers to estimation from low-degree polynomials

    Tselil Schramm and Alexander S. Wein. “Computational barriers to estimation from low-degree polynomials”. In:Ann. Statist.50.3 (2022), pp. 1833–1858.issn: 0090- 5364,2168-8966.doi:10.1214/22- aos2179.url:https://doi.org/10.1214/22- aos2179. 136 REFERENCES

    work page doi:10.1214/22- 2022

  33. [34]

    Sharp phase transitions in estimation with low-degree polynomials

    Youngtak Sohn and Alexander S. Wein. “Sharp phase transitions in estimation with low-degree polynomials”. In:STOC’25—Proceedings of the 57th Annual ACM Sympo- sium on Theory of Computing. ACM, New York, 2025, pp. 891–902.doi:10.1145/ 3717823.3718294.url:https://doi.org/10.1145/3717823.3718294

    work page doi:10.1145/3717823.3718294 2025

  34. [35]

    Stein and Rami Shakarchi.Complex analysis

    Elias M. Stein and Rami Shakarchi.Complex analysis. Vol. 2. Princeton Lectures in Analysis. Princeton University Press, Princeton, NJ, 2003, pp. xviii+379.isbn: 0-691- 11385-8

    work page 2003

  35. [36]

    Arabind Swain, Sean Alexander Ridout, and Ilya Nemenman.Better Together: Cross and Joint Covariances Enhance Signal Detectability in Undersampled Data. 2026. arXiv: 2507.22207 [cond-mat.dis-nn].url:https://arxiv.org/abs/2507.22207

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  36. [37]

    Hugo Tabanelli et al.Computational Thresholds in Multi-Modal Learning via the Spiked Matrix-Tensor Model. 2025. arXiv:2506.02664 [stat.ML].url:https://arxiv.org/ abs/2506.02664

    work page arXiv 2025

  37. [38]

    Terence Tao.Topics in random matrix theory. Vol. 132. Graduate Studies in Mathe- matics. American Mathematical Society, Providence, RI, 2012, pp. x+282.isbn: 978- 0-8218-7430-1.doi:10.1090/gsm/132.url:https://doi.org/10.1090/gsm/132

    work page doi:10.1090/gsm/132.url:https://doi.org/10.1090/gsm/132 2012

  38. [39]

    Solution of “Solvable model of a spin glass

    David J. Thouless, Philip W. Anderson, and Robert G. Palmer. “Solution of “Solvable model of a spin glass””. In:Philosophical Magazine35.3 (1977), pp. 593–601

    work page 1977

  39. [40]

    Christos Thrampoulidis, Samet Oymak, and Babak Hassibi.The Gaussian min-max theorem in the Presence of Convexity. 2015. arXiv:1408.4837 [cs.IT].url:https: //arxiv.org/abs/1408.4837

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  40. [41]

    Roman Vershynin.High-dimensional probability. Vol. 47. Cambridge Series in Statisti- cal and Probabilistic Mathematics. An introduction with applications in data science, With a foreword by Sara van de Geer. Cambridge University Press, Cambridge, 2018, pp. xiv+284.isbn: 978-1-108-41519-4.doi:10 . 1017 / 9781108231596.url:https : //doi.org/10.1017/9781108231596

    work page doi:10.1017/9781108231596 2018

  41. [42]

    Sharp thresholds for high-dimensional and noisy sparsity re- covery usingℓ 1-constrained quadratic programming (Lasso)

    Martin J. Wainwright. “Sharp thresholds for high-dimensional and noisy sparsity re- covery usingℓ 1-constrained quadratic programming (Lasso)”. In:IEEE Trans. Inform. Theory55.5 (2009), pp. 2183–2202.issn: 0018-9448,1557-9654.doi:10.1109/TIT. 2009.2016018.url:https://doi.org/10.1109/TIT.2009.2016018

    work page doi:10.1109/tit 2009

  42. [43]

    The Kikuchi hierarchy and tensor PCA

    Alexander S. Wein, Ahmed El Alaoui, and Cristopher Moore. “The Kikuchi hierarchy and tensor PCA”. In:2019 IEEE 60th Annual Symposium on Foundations of Computer Science. IEEE Comput. Soc. Press, Los Alamitos, CA, 2019, pp. 1446–1468.isbn: 978- 1-7281-4952-3

    work page 2019

  43. [44]

    Sharp Spectral Thresholds for Multi- View Spiked Wigner Models

    Yang Xiaodong, Subhabrata Sen, and Lu Yue M. “Sharp Spectral Thresholds for Multi- View Spiked Wigner Models”. In preparation. 2026

    work page 2026

  44. [45]

    Fundamental limits of community detection from multi-view data: multi-layer, dynamic and partially labeled block mod- els

    Xiaodong Yang, Buyu Lin, and Subhabrata Sen. “Fundamental limits of community detection from multi-view data: multi-layer, dynamic and partially labeled block mod- els”. In:Ann. Statist.53.6 (2025), pp. 2728–2756.issn: 0090-5364,2168-8966.doi: 10.1214/25-aos2562.url:https://doi.org/10.1214/25-aos2562

    work page doi:10.1214/25-aos2562.url:https://doi.org/10.1214/25-aos2562 2025

  45. [46]

    Statistical Physics of Inference: Thresholds and Algorithms

    Lenka Zdeborov´ a and Florent Krzakala. “Statistical Physics of Inference: Thresholds and Algorithms”. In:Advances in Physics65.5 (2016), pp. 453–552.doi:10.1080/ 00018732.2016.1211393. REFERENCES 137 Department of Mathematics, Massachusetts Institute of Technology Email address:hangdu@mit.edu Department of Mathematics, Massachusetts Institute of Technolo...

    work page arXiv 2016