Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

Strong replica symmetry for high-dimensional disordered log-concave Gibbs measures

T0 review · 3 major / 5 minor · reviewed 2026-08-27 · deepseek-v4-flash

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

Pith's one-line read For a generic class of random log-concave Gibbs measures, all multioverlaps self-average and the asymptotic Gibbs measure becomes a mixture of product measures.

desk verdict A real proof of multioverlap concentration for regularized log-concave Gibbs measures, but the advertised scope is wider than the theorem: the Gaussian tilt stays on and the free-entropy variance condition excludes simple linear exchangeable models. read the letter →

arxiv 2009.12939 v3 pith:JIKTWT7T submitted 2020-09-27 math.PR cond-mat.dis-nncs.ITcs.LGmath-phmath.ITmath.MP

Jean Barbier , Dmitry Panchenko , Manuel Sáenz This is my paper · ORCID
classification math.PRcond-mat.dis-nncs.ITcs.LGmath-phmath.ITmath.MP
keywords gibbsmeasuresinferencelog-concavehigh-dimensionalprovereplicastrong
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

Many high-dimensional systems are described by Gibbs measures, exponential weights exp(H) over many variables. When H is concave, the measure is log-concave and has strong concentration properties. This paper asks whether overlap parameters, which measure how two or more independent samples from the system agree, become deterministic in the large-system limit, a property called replica symmetry. The authors prove that, under a mild free-energy fluctuation condition, all such overlaps concentrate simultaneously. They first add a tiny quadratic (Gaussian) penalty to make the system strictly concave; this gives thermal concentration through the Brascamp-Lieb inequality. Then they add a Poisson-type random perturbation to force certain identities, use those identities to show all multioverlaps self-average, and finally remove the Poisson perturbation with an explicit error bound. The result implies that, asymptotically, the Gibbs measure becomes a mixture of product measures: coordinates behave nearly independently given a random environment. A caveat is that the main theorem is for the Gaussian-regularized measure; without assuming the original Hamiltonian is uniformly strictly concave, the regularization is not explicitly removed.
Extended reading notes

Core claim

Theorem 2: Under concavity of HN, spin exchangeability, and the free-entropy variance condition (v_N^{1/2}/s_N ->0), for every multioverlap R^(k), E< (R^(k)-E<R^(k)>_0)^2>_0 ->0, and consequently the asymptotic Gibbs measure has the product-form representation of Corollaries 2 and 3.

Load-bearing premise

The proof assumes the perturbed free entropy variance is O(N) and that a sequence s_N exists with s_N -> infinity, s_N/N -> 0, and (v'_N)^{1/2}/s_N -> 0 (defined in eq. (8) and Section 3). This is a disorder-dependent concentration condition on the log-partition function that is not derived from log-concavity or exchangeability; if free energy fluctuations grow faster than O(N), the Franz-de Sanctis bound (Theorem 5) and the decoupling lemma (Lemma 5) do not close. In addition, Theorem 2 is stated for the Gaussian-regularized measure < >_0, since the regularization is not removed for general concave Hamiltonians.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Request a human review

A listed scientist reviews the paper for a fee and the review publishes here regardless of verdict. See the reviewers or get listed.

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 studies high-dimensional Gibbs measures on [-1,1]^N with concave, possibly random Hamiltonians that are exchangeable under spin permutations. It introduces a vanishing Gaussian regularization and an auxiliary Poisson perturbation, and proves concentration of all multioverlaps: first thermal concentration via Brascamp-Lieb (Theorem 1), then, under a free-entropy variance condition and the Poisson perturbation, quenched concentration for the perturbed model (Proposition 1). Proposition 2 shows the Poisson perturbation affects mean multioverlaps by O((s_N/N)^{1/6}), and Theorem 2 is stated as strong replica symmetry for the Gaussian-regularized measure. Corollaries 2 and 3 derive an asymptotic product-form spin distribution and strong asymptotic independence via the Aldous-Hoover representation. The main theorem is proved for the regularized Gibbs measure, and it is conditional on a disorder-dependent concentration hypothesis on the (perturbed) free entropy.

Significance. If the theorem holds as stated, the paper gives a broad sufficient condition for replica-symmetric behavior in disordered log-concave systems, going beyond the Bayes-optimal setting of prior work [23] and beyond the two-replica overlap results of [38,35]. The proof is self-contained, gives explicit quantitative bounds for the perturbation removal, and uses a clean combination of Brascamp-Lieb, Efron-Stein, and Aldous-Hoover techniques. However, the advertised scope is wider than the proved result: Theorem 2 concerns the Gaussian-regularized measure, and the free-entropy variance hypothesis is a genuine additional concentration assumption that is not implied by concavity and exchangeability. These qualifications are important for applications to non-Bayes-optimal inference but do not, by themselves, invalidate the central derivation under the stated assumptions.

major comments (3)
  1. [Section 2.1 (Theorem 2) and Sections 3, 5.4] Theorem 2 is stated with the condition v_N^{1/2}/s_N -> 0 for the unperturbed regularized free-entropy variance v_N defined in (8), but its proof in Section 5.4 invokes Proposition 1, which requires the analogous condition (v'_N)^{1/2}/s_N -> 0 for the Poisson-perturbed variance v'_N defined in (18). No argument is given that v'_N can be controlled by v_N; the Poisson perturbation introduces additional randomness (pi, U, lambda) and a supremum over lambda, so the relation is not immediate. Consequently, Theorem 2 as stated is not proved. Please either add a transfer argument from v_N to v'_N, or restate Theorem 2 and Corollaries 2-3 with the v'_N condition, or prove Proposition 1 directly under the v_N hypothesis.
  2. [Abstract, Section 2.1, and Introduction] The main theorem is proved only for the Gaussian-regularized Gibbs measure < . >_0; the epsilon_N^2 ||sigma||^2 / 2 tilt is not removed for general concave Hamiltonians, as the authors themselves note before equation (5). This is a genuine scope restriction: the abstract's phrase "a generic class of log-concave, possibly random, (Gibbs) measures" and the title's unrestricted "log-concave Gibbs measures" overstate what is established. The introduction's claim that the results are established "for the original model whenever the Hamiltonian is uniformly strictly concave" is also stronger than Theorem 2, which does not state a uniform strict concavity condition. The regularized scope and the condition under which regularization can be omitted should be stated prominently in the abstract and in the theorem statements.
  3. [Section 2 (Perturbed free entropy variance)] The hypothesis v_N^{1/2}/s_N -> 0 is load-bearing and is not a consequence of concavity (2) or exchangeability (3). For example, take H_N(sigma | J) = J_N \sum_{i=1}^N sigma_i with J_N = sqrt(N) Z and Z ~ N(0,1). This Hamiltonian is twice differentiable, concave, and exchangeable, yet F_N = N log(2 sinh(J_N)/J_N) is asymptotic to N^{3/2}|Z| plus lower-order terms, so Var(F_N) is of order N^3 and no sequence s_N = o(N) can satisfy v_N^{1/2}/s_N -> 0. Thus the theorem does not apply to a model squarely inside the class suggested by the abstract. Calling this assumption "weak" in the paragraph after Theorem 2 is misleading; it is a substantive disorder-concentration condition that should be featured as such.
minor comments (5)
  1. [Section 2, equation (5)] There is a typo: "regularisation strenght" should be "regularisation strength".
  2. [Section 3, equation (18)] The sentence after (18) says E_J is expectation with respect to J only, with pi fixed, but the outer expectation in (18) is E over (J, pi, U). Please clarify which variables are averaged in each term of v'_N; the current wording is confusing.
  3. [Section 5.3, proof of Proposition 2] In the proof, the regularization strength is set to epsilon_N = (s_N/N)^{1/3}, but the hypotheses of Theorem 2 only say that epsilon_N satisfies epsilon_N -> 0 and N epsilon_N -> infinity. The dependence of epsilon_N on s_N should be made explicit in the statement of Theorem 2 and Corollaries 2 and 3, since the Gaussian regularization is part of the measure < . >_0.
  4. [Section 5.4] In the display for the triangular inequality, the notation E_lambda || <R> - E<R> ||_2 is not fully defined: the norm appears to be the L^2 norm over the Gibbs expectation and disorder, but the exact combination should be stated explicitly.
  5. [References] References [5] and [12] have incomplete author lists; for example, [5] lists no authors. Please complete them.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: multioverlap concentration is proved from concavity, exchangeability, and an explicit free-entropy variance hypothesis, not from itself.

full rationale

Walking the derivation chain: Theorem 1 is a direct application of the Brascamp-Lieb inequality to the Gaussian-regularized Hessian bound, so it is not circular. Proposition 1 is proved for the Poisson-perturbed model under the separate hypothesis (v'_N)^{1/2}/s_N -> 0 (Eq. 18 and Section 3); Lemma 3 bounds energy fluctuations, Theorem 5 converts this into a Franz-de Sanctis-type inequality, Lemma 5 decouples two replicas, and the induction shows every asymptotic multioverlap is a.s. constant, contradicting any assumed non-concentration. Proposition 2 removes the perturbation via a derivative bound together with Theorem 1, and Theorem 2 assembles these results. At no point is the target E<(R(k)-E<R(k)>_0)^2>_0 -> 0 used as an input. The variance condition is a disorder-dependent assumption on free-energy fluctuations; the paper explicitly labels it an assumption and notes it restricts the class to 'well-behaved' models. The self-citations [23] and [13] are contextual or auxiliary (e.g., for the o(N) free-energy comparison and a technical thermal-mean fact reproduced in the appendix), and the main proof is self-contained. The linear-model example in the skeptical reading shows a scope limitation, not circularity: it violates a hypothesis of the theorem. Hence no circular step can be exhibited.

Assumptions & free parameters 3 free parameters · 6 assumptions · 0 invented entities

The central claim depends on standard concentration and exchangeability tools, on the log-concavity and exchangeability of the Hamiltonian, and on an explicit free entropy variance condition. The Gaussian and Poisson perturbations are proof devices, not new physical entities. No new particles or forces are introduced.

free parameters (3)
  • Gaussian regularization strength epsilon_N = epsilon_N -> 0, N epsilon_N -> +infinity; in Proposition 2 set epsilon_N = (s_N/N)^{1/3}
    Introduced to ensure the Hessian is upper bounded by -epsilon_N I and to obtain thermal concentration via Brascamp-Lieb. The main theorem is for the regularized measure.
  • Poisson perturbation intensity s_N = s_N -> +infinity, s_N/N -> 0, chosen so that (v'_N)^{1/2}/s_N -> 0
    Controls the strength of the added Poisson disorder. It is needed for the Franz-de Sanctis type inequalities and is removed in Proposition 2.
  • Interpolation parameter t = t=1 for the main proof, t=0 for the unperturbed model
    Tunes the Poisson perturbation strength when bridging the perturbed and unperturbed models in Proposition 2.
assumptions (6)
  • standard math Brascamp-Lieb inequality and its simplified form (Theorem 3 and Corollary 4)
    Core variance bound used in Theorem 1 and Proposition 2.
  • standard math Aldous-Hoover representation for separately exchangeable arrays (Theorem 4)
    Used to represent subsequential limits of the spin arrays and to derive the product-form limiting description.
  • domain assumption The Hamiltonian is twice differentiable and concave for almost every disorder realization, with negative semidefinite Hessian (eq. (2))
    Defines the class of log-concave Gibbs measures studied.
  • domain assumption Spin exchangeability in distribution under permutations of spin indices (eq. (3))
    Needed for separate exchangeability of the spin array and hence for the Aldous-Hoover representation.
  • ad hoc to paper Free entropy variance condition: exists s_N with s_N -> infinity, s_N/N -> 0, and (v'_N)^{1/2}/s_N -> 0, with v'_N = O(N)
    Assumption on disorder fluctuations that is not proved from log-concavity and exchangeability. The central proof depends on it through Theorem 5 and Lemma 5.
  • domain assumption Spins are bounded in [-1,1] and can be rescaled without loss of generality
    Used for tightness and boundedness of multioverlaps throughout the proofs.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Strong replica symmetry for high-dimensional disordered log-concave Gibbs measures." pith.science (2026). https://pith.science/paper/JIKTWT7T

@misc{pith2026200912939,
  author       = {Pith},
  title        = {Pith review of: Strong replica symmetry for high-dimensional disordered log-concave Gibbs measures},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JIKTWT7T}},
  note         = {Machine review of arXiv:2009.12939}
}
read the original abstract

We consider a generic class of log-concave, possibly random, (Gibbs) measures. We prove the concentration of an infinite family of order parameters called multioverlaps. Because they completely parametrise the quenched Gibbs measure of the system, this implies a simple representation of the asymptotic Gibbs measures, as well as the decoupling of the variables in a strong sense. These results may prove themselves useful in several contexts. In particular in machine learning and high-dimensional inference, log-concave measures appear in convex empirical risk minimisation, maximum a-posteriori inference or M-estimation. We believe that they may be applicable in establishing some type of "replica symmetric formulas" for the free energy, inference or generalisation error in such settings.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Variational Bounds for Perceptron Learning from Structured Data

    cs.LG 2026-08 conditional novelty 7.0 of 10

    A variational potential gives both a lower and an upper bound on the quenched pressure of a perceptron, with matching supported numerically for logistic-type losses.

Reference graph

Works this paper leans on

66 extracted references · 66 canonical work pages · cited by 1 Pith paper

  1. [23]

    Strong replica symmetry in high-dimensional optimal Bayesian inference

    Barbier, J. & Panchenko, D. (2020) Strong replica symmetry in high-dimensional optima l Bayesian inference. arXiv preprint arXiv:2005.03115

    work page Pith review arXiv 2020

  2. [13]

    Performance of Bayesian linear regression in a model with mismatch

    Barbier, J., Chen, W.-K., Panchenko, D. & S´aenz, M. (2021) Performance of Bayesian linear regression in a model with mismatch. arXiv preprint arXiv:2107.06936

    work page Pith review arXiv 2021

  3. [1]

    & Tomczak- Jaegermann, N

    Adamczak, R., Lata/suppress la, R., Litvak, A., Oleszkiewicz, K., Pajor,A. & Tomczak- Jaegermann, N. (2014) A short proof of Paouris’ inequality. Canadian Mathematical Bulletin , 57(1), 3–8

    work page 2014

  4. [2]

    & Contucci, P

    Aizenman, M. & Contucci, P. (1998) On the stability of the quenched state in mean-field spin-glass models. Journal of Statistical Physics , 92(5-6), 765–783

    work page 1998

  5. [3]

    & Starr, S

    Aizenman, M., Sims, R. & Starr, S. L. (2003) Extended variational principle for the Sherrington-Kirkpatrick spin-glass model. Physical Review B , 68(21), 214403

    work page 2003

  6. [4]

    The solution of the deep Boltzmann machine on the Nishimori line

    Alberici, D., Camilli, F., Contucci, P. & Mingione, E. (2020) The solution of the deep Boltzmann machine on the Nishimori line. arXiv preprint arXiv:2012.13987 . 23

    work page Pith review arXiv 2020

  7. [5]

    Journal of Statistical Physics , 182(1), 1–20

    (2021) The Multi-species Mean-Field Spin-Glass on the Nish imori Line. Journal of Statistical Physics , 182(1), 1–20

    work page 2021

  8. [6]

    (1981) Representations for partially exchangeable arrays of random variables

    Aldous, D. (1981) Representations for partially exchangeable arrays of random variables. Journal of Multivariate Analysis , 11(4), 581–598

    work page 1981

Show all 66 references
  1. [7]

    C., Ricci-Tersenghi, F

    Angelini, M. C., Ricci-Tersenghi, F. & Kabashima, Y. (2012) Compressed sensing with sparse, structured matrices. in 2012 50th Annual Allerton Conference on Communication, Con- trol, and Computing (Allerton) , pp. 808–814. IEEE

    2012

  2. [8]

    & Zdeborov´a, L

    Aubin, B., Maillard, A., Barbier, J., Krzakala, F., Macris, N. & Zdeborov´a, L. (2018) The committee machine: computational to statistica l gaps in learning a two-layers neural network. in Advances in Neural Information Processing Systems , pp. 3223–3234

    2018

  3. [9]

    & Coja-Oghlan, A

    Bapst, V. & Coja-Oghlan, A. (2016) Harnessing the Bethe free energy. Random Structures & Algorithms , 49(4), 694–741

    2016

  4. [10]

    (2020) Overlap matrix concentration in optimal Bayesian in ference

    Barbier, J. (2020) Overlap matrix concentration in optimal Bayesian in ference. Information and Inference: a Journal of the IMA , iaaa008

    2020

  5. [11]

    & Macris, N

    Barbier, J., Chan, C. & Macris, N. (2018) Adaptive path interpolation for sparse sys- tems: application to a simple censored block model. in 2018 IEEE International Symposium on Information Theory (ISIT) , pp. 1879–1883. IEEE

    2018

  6. [12]

    Journal of Statistical Physics , 180, 534–557

    (2020) Concentration of multi-overlaps for random ferroma gnetic spin models. Journal of Statistical Physics , 180, 534–557

    2020

  7. [14]

    & Zdeborov´a, L

    Barbier, J., Krzakala, F., Macris, N., Miolane, L. & Zdeborov´a, L. (2019) Optimal errors and phase transitions in high-dimensional generali zed linear models. Proceedings of the National Academy of Sciences , 116(12), 5451–5460

    2019

  8. [15]

    & Macris, N

    Barbier, J., Luneau, C. & Macris, N. (2020) Mutual information for low-rank even-order symmetric tensor factorization. Information and Inference: a Journal of the IMA , iaaa022

    2020

  9. [16]

    & Macris, N

    Barbier, J. & Macris, N. (2019a) 0-1 phase transitions in sparse spiked matrix estim ation. arXiv preprint arXiv:1911.05030

    2019 arXiv

  10. [17]

    Probability Theory and Related Fields , 174(3-4), 1133–1185

    (2019b) The adaptive interpolation method: a simple scheme to prove replica formulas in Bayesian inference. Probability Theory and Related Fields , 174(3-4), 1133–1185

    2019

  11. [18]

    Applica- tions to the Curie–Weiss and Wigner spike models

    (2019c) The adaptive interpolation method for proving repl ica formulas. Applica- tions to the Curie–Weiss and Wigner spike models. Journal of Physics A: Mathematical and Theoretical, 52(29), 294002

    2019

  12. [19]

    & Krzakala, F

    Barbier, J., Macris, N., Dia, M. & Krzakala, F. (2020) Mutual information and op- timality of approximate message-passing in random linear e stimation. IEEE Transactions on Information Theory, 66(7), 4270–4303

    2020

  13. [20]

    & Krzakala, F

    Barbier, J., Macris, N., Maillard, A. & Krzakala, F. (2018) The mutual information in random linear estimation beyond iid matrices. in 2018 IEEE International Symposium on Information Theory (ISIT) , pp. 1390–1394. IEEE. 24

    2018

  14. [21]

    & Miolane, L

    Barbier, J., Macris, N. & Miolane, L. (2017) The layered structure of tensor estima- tion and its mutual information. in 2017 55th Annual Allerton Conference on Communication, Control, and Computing (Allerton) , pp. 1056–1063. IEEE

    2017

  15. [22]

    & Rush, C

    Barbier, J., Macris, N. & Rush, C. (2020) All-or-nothing statistical and computational phase transitions in sparse spiked matrix estimation. in Advances in Neural Information Pro- cessing Systems

    2020

  16. [24]

    & De Sanctis, L

    Barra, A. & De Sanctis, L. (2007) Stability properties and probability distribution s of multi-overlaps in dilute spin glasses. Journal of Statistical Mechanics: Theory and Experiment , 2007(08), P08025

    2007

  17. [25]

    (1984) Convexity of measures in certain convex cones in vect or space σ-algebras

    Borell, C. (1984) Convexity of measures in certain convex cones in vect or space σ-algebras. Mathematica Scandinavica, 53(1), 125–144

    1984

  18. [26]

    & Vandenberghe, L

    Boyd, S. & Vandenberghe, L. (2004) Convex optimization . Cambridge University Press

    2004

  19. [27]

    & Lieb, E

    Brascamp, H. & Lieb, E. (2002) On extensions of the Brunn-Minkowski and Pr´ ekopa- Leindler theorems, including inequalities for log concave functions, and with an application to the diffusion equation. in Inequalities, pp. 441–464. Springer

    2002

  20. [28]

    (2015) Absence of replica symmetry breaking in the random fie ld Ising model

    Chatterjee, S. (2015) Absence of replica symmetry breaking in the random fie ld Ising model. Communications in Mathematical Physics , 337(1), 93–102

    2015

  21. [29]

    & Ravelomanana, J

    Coja-Oghlan, A., Cooley, O., Kang, M., Lee, J. & Ravelomanana, J. B. (2021) The sparse parity matrix. arXiv preprint arXiv:2107.06123

    2021 arXiv

  22. [30]

    & Kapetanopoulos, T

    Coja-Oghlan, A., Efthymiou, C., Jaafari, N., Kang, M. & Kapetanopoulos, T. (2018) Charting the replica symmetric phase. Communications in Mathematical Physics , 359(2), 603–698

    2018

  23. [31]

    & M¨uller, N

    Coja-Oghlan, A., Kapetanopoulos, T. & M¨uller, N. (2018) The replica symmetric phase of random constraint satisfaction problems. Combinatorics, Probability and Computing , pp. 1–77

    2018

  24. [32]

    & Zdeborov´a, L

    Coja-Oghlan, A., Krzakala, F., Perkins, W. & Zdeborov´a, L. (2018) Information- theoretic thresholds from the cavity method. Advances in Mathematics , 333, 694–795

    2018

  25. [33]

    & Nishimori, H

    Contucci, P., Giardina, C. & Nishimori, H. (2009) Spin glass identities and the Nishimori line. in Spin Glasses: Statics and Dynamics , pp. 103–121. Springer

    2009

  26. [34]

    (2004) Random multi-overlap structures and cavity fields in diluted spin glasses

    De Sanctis, L. (2004) Random multi-overlap structures and cavity fields in diluted spin glasses. Journal of statistical physics , 117(5), 785–799

    2004

  27. [35]

    & Franz, S

    De Sanctis, L. & Franz, S. (2009) Self-averaging identities for random spin systems. in Spin glasses: statics and dynamics , pp. 123–142. Springer

    2009

  28. [36]

    Dia, M., Macris, N., Krzakala, F., Lesieur, T., Zdeborov ´a, L. et al. (2016) Mutual information for symmetric rank-one matrix estimation: A pr oof of the replica formula. Advances in Neural Information Processing Systems , 29, 424–432. 25

    2016

  29. [37]

    & Krzakala, F

    El Alaoui, A. & Krzakala, F. (2018) Estimation in the spiked Wigner model: a short proof of the replica formula. in 2018 IEEE International Symposium on Information Theory (ISIT), pp. 1874–1878. IEEE

    2018

  30. [38]

    & Leone, M

    Franz, S. & Leone, M. (2003) Replica bounds for optimization problems and dilute d spin systems. Journal of Statistical Physics , 111(3-4), 535–564

    2003

  31. [39]

    & Zdeborov´a, L

    Gabri´e, M., Manoel, A., Luneau, C., Barbier, J., Macris, N., Krzakala, F. & Zdeborov´a, L. (2018) Entropy and mutual information in models of deep neur al networks. in Advances in Neural Information Processing Systems , pp. 1821–1831

    2018

  32. [40]

    & Guerra, F

    Ghirlanda, S. & Guerra, F. (1998) General properties of overlap probability distribu tions in disordered spin systems: towards Parisi ultrametricity . Journal of Physics A: Mathematical and General, 31(46), 9149

    1998

  33. [41]

    & Urbanke, R

    Giurgiu, A., Macris, N. & Urbanke, R. (2016) Spatial coupling as a proof technique and three applications. IEEE Transactions on Information Theory , 62(10), 5281–5295

    2016

  34. [42]

    (2003) Broken replica symmetry bounds in the mean field spin g lass model

    Guerra, F. (2003) Broken replica symmetry bounds in the mean field spin g lass model. Communications in Mathematical Physics , 233(1), 1–12

    2003

  35. [43]

    & Toninelli, F

    Guerra, F. & Toninelli, F. (2002) The thermodynamic limit in mean field spin glass models. Communications in Mathematical Physics , 230(1), 71–79

    2002

  36. [44]

    & Toninelli, F

    Guerra, F. & Toninelli, F. L. (2004) The high temperature region of the Viana–Bray diluted spin glass model. Journal of statistical physics , 115(1), 531–555

    2004

  37. [45]

    (1982) Row-column exchangeability and a generalized model for probability

    Hoover, D. (1982) Row-column exchangeability and a generalized model for probability. Exchangeability in Probability and Statistics (Rome, 1981 ), pp. 281–291

    1982

  38. [46]

    & Ronchetti, E

    Huber, P. & Ronchetti, E. (2009) Robust statistics, Wiley Series in Probability and Statis- tics. Wiley

    2009

  39. [47]

    & Miolane, L

    Lelarge, M. & Miolane, L. (2018) Fundamental limits of symmetric low-rank matrix estimation. Probabability Theory and Related Fields , 173(3-4), 859–929

    2018

  40. [48]

    & Zdeborov´a, L

    Lesieur, T., Miolane, L., Lelarge, M., Krzakala, F. & Zdeborov´a, L. (2017) Sta- tistical and computational phase transitions in spiked ten sor estimation. in 2017 IEEE Inter- national Symposium on Information Theory (ISIT) , pp. 511–515. IEEE

    2017

  41. [49]

    & Macris, N

    Luneau, C., Barbier, J. & Macris, N. (2020) Information theoretic limits of learning a sparse rule. in Advances in Neural Information Processing Systems

    2020

  42. [50]

    & Montanari, A

    M´ezard, M. & Montanari, A. (2009) Information, physics, and computation . Oxford Uni- versity Press

    2009

  43. [51]

    & Virasoro, M

    M´ezard, M., Parisi, G. & Virasoro, M. (1987) Spin glass theory and beyond: an intro- duction to the replica method and its applications , vol. 9. World Scientific Publishing Company

    1987

  44. [52]

    & Shah, D

    Montanari, A. & Shah, D. (2006) Counting good truth assignments of random k-SAT formulae. arXiv preprint cs/0607073

    2006 arXiv

  45. [53]

    (2001) Statistical physics of spin glasses and information proces sing: an intro- duction

    Nishimori, H. (2001) Statistical physics of spin glasses and information proces sing: an intro- duction. Clarendon Press. 26

    2001

  46. [54]

    (2010) The Ghirlanda–Guerra identities for mixed p-spin model

    Panchenko, D. (2010) The Ghirlanda–Guerra identities for mixed p-spin model. Comptes Rendus Mathematique , 348(3-4), 189–192

    2010

  47. [55]

    Springer Science & Business Media

    (2013a) The Sherrington-Kirkpatrick model . Springer Science & Business Media

    2013

  48. [56]

    The Annals of Probability, 41(3A), 1315–1361

    (2013b) Spin glass models from the point of view of spin distr ibutions. The Annals of Probability, 41(3A), 1315–1361

    2013

  49. [57]

    Electronic Journal of Probability, 19

    (2014) On the replica symmetric solution of the K-sat model. Electronic Journal of Probability, 19

    2014

  50. [58]

    Journal of Statistical Physics , 162(1), 1–42

    (2016) Structure of finite-RSB asymptotic Gibbs measures in the diluted spin glass models. Journal of Statistical Physics , 162(1), 1–42

    2016

  51. [59]

    (2006) Concentration of mass on convex bodies

    Paouris, G. (2006) Concentration of mass on convex bodies. Geometric and Functional Anal- ysis, 16, 1021–1049

    2006

  52. [60]

    (2020) Information-theoretic limits for the matrix tensor product

    Reeves, G. (2020) Information-theoretic limits for the matrix tensor product. IEEE Journal on Selected Areas in Information Theory

    2020

  53. [61]

    & Pfister, H

    Reeves, G. & Pfister, H. D. (2016) The replica-symmetric prediction for compressed sensing with Gaussian matrices is exact. in 2016 IEEE International Symposium on Information Theory (ISIT) , pp. 665–669. IEEE

    2016

  54. [62]

    & Wellner, J

    Saumard, A. & Wellner, J. (2014) Log-concavity and strong log-concavity: a review. Statistics Surveys , 8, 45

    2014

  55. [63]

    & Barra, A

    Sollich, P. & Barra, A. (2012) Spin glass polynomial identities from entropic cons traints. Journal of Physics A: Mathematical and Theoretical , 45(48), 485001

    2012

  56. [64]

    (2001) The high temperature case for the random K-sat problem

    Talagrand, M. (2001) The high temperature case for the random K-sat problem. Probability Theory and Related Fields , 119(2), 187–212

    2001

  57. [65]

    (2010) Mean field models for spin glasses: Volume I: Basic examples , vol. 54. Springer Science & Business Media

    2010

  58. [66]

    Volume II: advanced replica -symmetry and low temperature

    (2011) Mean field models for spin glasses. Volume II: advanced replica -symmetry and low temperature. Springer Verlag. 27

    2011

Pith tools

Reviewed August 27, 2026 · model on record in the stance chip above.