pith. sign in

arxiv: 1907.01823 · v2 · pith:QOHG6OFXnew · submitted 2019-07-03 · 🧮 math.FA · math.PR

Spectral gaps, symmetries and log-concave perturbations

Franck Barthe , Bo'az Klartag This is my paper

Pith reviewed 2026-05-25 09:56 UTC · model grok-4.3

classification 🧮 math.FA math.PR
keywords Poincaré constantlog-concave measuresspectral gapsymmetric exponential measuresections of convex bodiesell_p ballseigenvalue multiplicityspectrum interlacing
0
0 comments X

The pith

Even log-concave perturbations of the symmetric exponential measure keep Poincaré constants growing at most logarithmically in dimension.

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

The paper studies how log-concave perturbations affect the spectral gap, or Poincaré constant, of a probability measure on R^n. For the symmetric exponential measure, any even log-concave perturbation produces a Poincaré constant that increases at most like the logarithm of the dimension. The same bound supplies estimates, optimal up to logarithmic factors, for the Poincaré constants of (n/2)-dimensional sections of the unit ball of ell_p^n when 1 ≤ p ≤ 2. Under symmetry assumptions the associated Laplace-type operator has an eigenspace of exact dimension n for its lowest positive eigenvalue, and the odd and even parts of the spectrum interlace.

Core claim

Perturbing the symmetric exponential measure by an even log-concave density does not increase its Poincaré constant by more than a logarithmic factor in the dimension n. This yields Poincaré-constant bounds for (n/2)-dimensional sections of the ell_p^n unit ball (1 ≤ p ≤ 2) that are optimal up to log factors. Under symmetry assumptions the eigenspace of the lowest positive eigenvalue of the associated Laplace-type operator has dimension exactly n, and the odd and even parts of the spectrum interlace.

What carries the argument

The Poincaré constant of an even log-concave perturbation of the symmetric exponential measure, controlled by the base measure's own spectral properties.

If this is right

  • Poincaré constants of (n/2)-dimensional sections of the ell_p^n unit ball (1 ≤ p ≤ 2) remain bounded by a multiple of log n.
  • The lowest positive eigenvalue of the Laplace-type operator has multiplicity exactly n under the stated symmetry assumptions.
  • The odd and even parts of the spectrum satisfy an interlacing relation.
  • The logarithmic growth is inherited by many high-dimensional convex bodies obtained as sections or projections.

Where Pith is reading between the lines

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

  • Similar logarithmic control may hold when the base measure is replaced by other product measures with sub-exponential tails.
  • The symmetry-based dimension count could be used to obtain sharp bounds on higher eigenvalues for symmetric log-concave measures.
  • The section estimates suggest that functional inequalities for ell_p balls remain stable under random hyperplane sections up to log factors.

Load-bearing premise

The perturbations must be both log-concave and even, and the base measure must be the symmetric exponential distribution.

What would settle it

An explicit even log-concave perturbation of the symmetric exponential measure in high dimension whose Poincaré constant grows faster than C log n for every constant C would disprove the main bound.

read the original abstract

We discuss situations where perturbing a probability measure on $\mathbb{R}^n$ does not deteriorate its Poincar\'e constant by much. A particular example is the symmetric exponential measure in $\mathbb{R}^n$, even log-concave perturbations of which have Poincar\'e constants that grow at most logarithmically with the dimension. This leads to estimates for the Poincar\'e constants of $(n/2)$-dimensional sections of the unit ball of $\ell_p^n$ for $1 \leq p \leq 2$, which are optimal up to logarithmic factors. We also consider symmetry properties of the eigenspace of the Laplace-type operator associated with a log-concave measure. Under symmetry assumptions we show that the dimension of this space is exactly $n$, and we exhibit a certain interlacing between the "odd" and "even" parts of the spectrum.

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

2 major / 2 minor

Summary. The paper establishes that even log-concave perturbations of the symmetric exponential measure on R^n have Poincaré constants that grow at most logarithmically in the dimension. It applies this to derive bounds on the Poincaré constants of (n/2)-dimensional central sections of the unit ball of ℓ_p^n (1 ≤ p ≤ 2) that are optimal up to logarithmic factors. It further shows that, under suitable symmetry assumptions on a log-concave measure, the eigenspace of the associated Laplace-type operator has dimension exactly n and that the odd and even parts of the spectrum interlace.

Significance. If the results hold, the perturbation theorem supplies a concrete mechanism for controlling spectral gaps of high-dimensional log-concave measures relative to the exponential distribution, which is a useful addition to the literature on Poincaré inequalities. The application yields new geometric estimates for sections of ℓ_p balls that connect functional-analytic bounds to questions in asymptotic convex geometry. The symmetry results on the spectrum are of independent interest for the spectral theory of log-concave measures. The manuscript states its theorems clearly and provides explicit constants in the main estimates.

major comments (2)
  1. [Application to sections of ℓ_p balls] The section deriving the bounds for (n/2)-dimensional sections of the ℓ_p unit ball (the part following the perturbation theorem): the claim that these uniform measures arise as even log-concave perturbations of the symmetric exponential measure μ without introducing extra dimension-dependent factors is load-bearing for the optimality statement, yet the explicit representation, verification that the resulting density remains even and log-concave, and control of the Jacobian under the restriction to the section are not supplied in sufficient detail to confirm that the C log n bound survives unchanged.
  2. [Symmetry properties of the eigenspace] § on symmetry properties of the eigenspace: the statement that the dimension is exactly n under the stated symmetry assumptions on the measure relies on the precise definition of the symmetry group and the decomposition into odd/even functions; the argument should explicitly identify the symmetry hypotheses that guarantee the eigenspace dimension equals n rather than being at most n.
minor comments (2)
  1. [Abstract] The abstract refers to 'estimates ... which are optimal up to logarithmic factors' without indicating the precise comparison (e.g., to the known lower bounds or to the exponential case itself); a brief parenthetical clarification would improve readability.
  2. [Notation and setup] Notation for the symmetric exponential measure and the perturbation density should be introduced once and used consistently; occasional shifts between μ and the perturbed measure make some displayed inequalities harder to parse.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading of the manuscript and the constructive comments. We address each major comment below. We agree that additional details and clarifications are needed and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [Application to sections of ℓ_p balls] The section deriving the bounds for (n/2)-dimensional sections of the ℓ_p unit ball (the part following the perturbation theorem): the claim that these uniform measures arise as even log-concave perturbations of the symmetric exponential measure μ without introducing extra dimension-dependent factors is load-bearing for the optimality statement, yet the explicit representation, verification that the resulting density remains even and log-concave, and control of the Jacobian under the restriction to the section are not supplied in sufficient detail to confirm that the C log n bound survives unchanged.

    Authors: We agree that the current presentation lacks sufficient explicit detail on this point. In the revised manuscript we will supply the explicit representation of the uniform measure on the central section as an even log-concave perturbation of the symmetric exponential measure, verify evenness and log-concavity of the density, and give a direct computation of the Jacobian under the restriction map showing that it contributes no additional dimension-dependent factors beyond the logarithmic term already present. revision: yes

  2. Referee: [Symmetry properties of the eigenspace] § on symmetry properties of the eigenspace: the statement that the dimension is exactly n under the stated symmetry assumptions on the measure relies on the precise definition of the symmetry group and the decomposition into odd/even functions; the argument should explicitly identify the symmetry hypotheses that guarantee the eigenspace dimension equals n rather than being at most n.

    Authors: We accept the referee’s observation. The existing argument shows that the dimension is at most n under the stated symmetries. In the revision we will explicitly list the minimal symmetry hypotheses (invariance under coordinate sign flips together with a transitive permutation action on the coordinates) and prove that these force the eigenspace to be spanned exactly by the n coordinate functions, thereby establishing dimension precisely n rather than at most n. We will also clarify the odd/even decomposition under these symmetries. revision: yes

Circularity Check

0 steps flagged

No circularity; self-contained theoretical derivation

full rationale

The paper presents a functional-analytic derivation of Poincaré constant bounds for even log-concave perturbations of the symmetric exponential measure, followed by an application to sections of ℓ_p balls. No self-definitional reductions, fitted inputs renamed as predictions, or load-bearing self-citation chains appear in the abstract or described claims. The central results are established via direct mathematical arguments on the Laplace-type operator and symmetry properties, without any quoted step that reduces by construction to its own inputs or prior author work. This is the expected outcome for a pure theory paper in math.FA.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The work rests on standard background facts from convex geometry and functional analysis (log-concavity implies certain concentration, existence of the Laplace operator for log-concave measures). No free parameters or invented entities are introduced in the abstract.

axioms (2)
  • domain assumption Log-concave measures on R^n admit a well-defined Laplace-type operator whose spectrum controls the Poincaré constant.
    Invoked throughout the abstract when discussing the operator and its eigenspace.
  • domain assumption Even symmetry of the perturbation preserves key properties of the base exponential measure.
    Central to the perturbation result stated in the abstract.

pith-pipeline@v0.9.0 · 5673 in / 1293 out tokens · 26840 ms · 2026-05-25T09:56:45.125220+00:00 · methodology

discussion (0)

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

Forward citations

Cited by 1 Pith paper

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

  1. Fast Mixing of Data Augmentation Algorithms: Bayesian Probit, Logit, and Lasso Regression

    math.ST 2024-12 unverdicted novelty 8.0

    The paper proves polynomial mixing time upper bounds for three data augmentation algorithms (ProbitDA, LogitDA, LassoDA) with explicit dependence on design matrix, prior, n, and d.

Reference graph

Works this paper leans on

34 extracted references · 34 canonical work pages · cited by 1 Pith paper · 1 internal anchor

  1. [1]

    Approaching the Kannan-Lov´ asz- Simonovits and variance conjectures , volume 2131 of Lecture Notes in Mathematics

    David Alonso-Guti´ errez and Jes´ us Bastero. Approaching the Kannan-Lov´ asz- Simonovits and variance conjectures , volume 2131 of Lecture Notes in Mathematics . Springer, Cham, 2015

    work page 2015

  2. [2]

    Analysis and geometry of Markov diffusion operators , volume 348 of Grundlehren der Mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences]

    Dominique Bakry, Ivan Gentil, and Michel Ledoux. Analysis and geometry of Markov diffusion operators , volume 348 of Grundlehren der Mathematischen Wissenschaften [Fundamental Principles of Mathematical Sciences] . Springer, Cham, 2014

    work page 2014

  3. [3]

    Barthe and D

    F. Barthe and D. Cordero-Erausquin. Invariances in variance e stimates. Proc. Lond. Math. Soc. (3) , 106(1):33–64, 2013

    work page 2013

  4. [4]

    Barthe and P

    F. Barthe and P. Wolff. Remarks on non-interacting conservativ e spin systems: the case of gamma distributions. Stochastic Process. Appl. , 119(8):2711–2723, 2009

    work page 2009

  5. [5]

    Bobkov and M

    S. Bobkov and M. Ledoux. Poincar´ e’s inequalities and Talagrand’s concentration phe- nomenon for the exponential distribution. Probab. Theory Related Fields , 107(3):383– 400, 1997

    work page 1997

  6. [6]

    S. G. Bobkov. Isoperimetric and analytic inequalities for log-conc ave probability mea- sures. Ann. Probab., 27(4):1903–1921, 1999

    work page 1903

  7. [7]

    S. G. Bobkov and M. Ledoux. From Brunn-Minkowski to Brascam p-Lieb and to loga- rithmic Sobolev inequalities. Geom. Funct. Anal. , 10(5):1028–1052, 2000

    work page 2000

  8. [8]

    Herm Jan Brascamp and Elliott H. Lieb. On extensions of the Brunn -Minkowski and Pr´ ekopa-Leindler theorems, including inequalities for log concave f unctions, and with an application to the diffusion equation. J. Functional Analysis , 22(4):366–389, 1976. 26

    work page 1976

  9. [9]

    Geometry of isotropic convex bodies , volume 196 of Mathematical Surveys and Monographs

    Silouanos Brazitikos, Apostolos Giannopoulos, Petros Valettas, and Beatrice-Helen Vritsiou. Geometry of isotropic convex bodies , volume 196 of Mathematical Surveys and Monographs. American Mathematical Society, Providence, RI, 2014

    work page 2014

  10. [10]

    Statistics for high-dimensional data

    Peter B¨ uhlmann and Sara van de Geer. Statistics for high-dimensional data . Springer Series in Statistics. Springer, Heidelberg, 2011. Methods, theory and applications

    work page 2011

  11. [11]

    Cordero-Erausquin, M

    D. Cordero-Erausquin, M. Fradelizi, and B. Maurey. The (B) co njecture for the Gaus- sian measure of dilates of symmetric convex sets and related proble ms. J. Funct. Anal. , 214(2):410–427, 2004

    work page 2004

  12. [12]

    Thin shell implies spectral gap up to polylog via a stoc hastic localization scheme

    Ronen Eldan. Thin shell implies spectral gap up to polylog via a stoc hastic localization scheme. Geom. Funct. Anal. , 23(2):532–569, 2013

    work page 2013

  13. [13]

    Gauss ian mixtures: entropy and geometric inequalities

    Alexandros Eskenazis, Piotr Nayar, and Tomasz Tkocz. Gauss ian mixtures: entropy and geometric inequalities. Ann. Probab., 46(5):2908–2945, 2018

    work page 2018

  14. [14]

    Lawrence C. Evans. Partial differential equations , volume 19 of Graduate Studies in Mathematics. American Mathematical Society, Providence, RI, 1998

    work page 1998

  15. [15]

    Spectral gaps for spin system s: some non-convex phase examples

    Ivan Gentil and Cyril Roberto. Spectral gaps for spin system s: some non-convex phase examples. J. Funct. Anal. , 180(1):66–84, 2001

    work page 2001

  16. [16]

    A convex/log-concave correlation inequality for G aussian measure and an application to abstract Wiener spaces

    Gilles Harg´ e. A convex/log-concave correlation inequality for G aussian measure and an application to abstract Wiener spaces. Probab. Theory Related Fields , 130(3):415–440, 2004

    work page 2004

  17. [17]

    Remarks on decay of correlations and Witten L aplacians

    Bernard Helffer. Remarks on decay of correlations and Witten L aplacians. III. Ap- plication to logarithmic Sobolev inequalities. Ann. Inst. H. Poincar´ e Probab. Statist., 35(4):483–508, 1999

    work page 1999

  18. [18]

    Spectral gap for some invariant log-concave pro bability measures

    Nolwen Huet. Spectral gap for some invariant log-concave pro bability measures. Math- ematika, 57(1):51–62, 2011

    work page 2011

  19. [19]

    Kannan, L

    R. Kannan, L. Lov´ asz, and M. Simonovits. Isoperimetric prob lems for convex bodies and a localization lemma. Discrete Comput. Geom. , 13(3-4):541–559, 1995

    work page 1995

  20. [20]

    Unimodality and dominance for symmetric random vectors

    Marek Kanter. Unimodality and dominance for symmetric random vectors. Trans. Amer. Math. Soc. , 229:65–85, 1977

    work page 1977

  21. [21]

    A Berry-Esseen type inequality for convex bodie s with an unconditional basis

    Bo’az Klartag. A Berry-Esseen type inequality for convex bodie s with an unconditional basis. Probab. Theory Related Fields , 145(1-2):1–33, 2009

    work page 2009

  22. [22]

    Kolesnikov and Emanuel Milman

    Alexander V. Kolesnikov and Emanuel Milman. The KLS isoperimetr ic conjecture for generalized Orlicz balls. Ann. Probab., 46(6):3578–3615, 2018

    work page 2018

  23. [23]

    Weighted Sobolev spaces

    Alois Kufner. Weighted Sobolev spaces . A Wiley-Interscience Publication. John Wiley & Sons, Inc., New York, 1985. Translated from the Czech

    work page 1985

  24. [24]

    Lata/suppress la and J

    R. Lata/suppress la and J. O. Wojtaszczyk. On the infimum convolution inequality. Studia Math., 189(2):147–187, 2008

    work page 2008

  25. [25]

    Eldan's Stochastic Localization and the KLS Conjecture: Isoperimetry, Concentration and Mixing

    Yin Tat Lee and Santosh Vempala. Eldan’s stochastic localization a nd the KLS hyper- plane conjecture: An improved lower bound for expansion. ArXiv 1612.01507, 2016

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  26. [26]

    Sections of the unit ball of Ln p

    Mathieu Meyer and Alain Pajor. Sections of the unit ball of Ln p . J. Funct. Anal. , 80(1):109–123, 1988

    work page 1988

  27. [27]

    On the role of convexity in isoperimetry, spectr al gap and concen- tration

    Emanuel Milman. On the role of convexity in isoperimetry, spectr al gap and concen- tration. Invent. Math. , 177(1):1–43, 2009

    work page 2009

  28. [28]

    V. D. Milman and A. Pajor. Isotropic position and inertia ellipsoids a nd zonoids of the unit ball of a normed n-dimensional space. In Geometric aspects of functional analysis (1987–88), volume 1376 of Lecture Notes in Math. , pages 64–104. Springer, Berlin, 1989

    work page 1987

  29. [29]

    Hardy’s inequality with weights

    Benjamin Muckenhoupt. Hardy’s inequality with weights. Studia Math. , 44:31–38,

    work page

  30. [30]

    Collection of articles honoring the completion by Antoni Zygmu nd of 50 years of scientific activity, I. 27

    work page

  31. [31]

    Tyrrell Rockafellar

    R. Tyrrell Rockafellar. Convex analysis. Princeton Landmarks in Mathematics. Prince- ton University Press, Princeton, NJ, 1997. Reprint of the 1970 or iginal, Princeton Paperbacks

    work page 1997

  32. [32]

    Poinca r´ e inequalities on intervals—application to sensitivity analysis

    Olivier Roustant, Franck Barthe, and Bertrand Iooss. Poinca r´ e inequalities on intervals—application to sensitivity analysis. Electron. J. Stat. , 11(2):3081–3119, 2017

    work page 2017

  33. [33]

    A simple proof of the Gaussian correlation conje cture extended to some multivariate gamma distributions

    Thomas Royen. A simple proof of the Gaussian correlation conje cture extended to some multivariate gamma distributions. Far East J. Theor. Stat. , 48(2):139–145, 2014

    work page 2014

  34. [34]

    An isoperimetric inequality on the lp balls

    Sasha Sodin. An isoperimetric inequality on the lp balls. Ann. Inst. Henri Poincar´ e Probab. Stat., 44(2):362–373, 2008. Institut de Math´ eatiques de Toulouse, CNRS UMR 5219, Universit´ e Paul Sabatier, 31062 Toulouse Cedex 09, France. barthe@math.univ-toulouse.fr Department of Mathematics, Weizmann Institute of Science, Reho vot 76100 Israel. boaz.klar...

    work page 2008