betaFeedback

WorksMathematics

Universality of Annealed Complexity in the Spherical Pure pp-Spin Model

Abstract

In an influential work, Auffinger, Ben Arous, and Černý (2013) investigated the limiting annealed complexity in the spherical pure pp-spin model associated to the Gaussian disorder. Their results, in particular, included an explicit expression for the total number of critical points of the Hamiltonian below any given energy level. In this paper, we consider the same anneal complexity associated to a disorder interpolating between an arbitrary random variable and its Gaussian counterpart. Assuming strict subgaussianity of the former, our result shows that the annealed complexity shares the same limit as in the Gaussian case.

Introduction and Main Results

The study of the complexity in the spin glass theory is concerned about the total numbers of critical or saddle points for the disordered Hamiltonians. By analyzing their asymptotic behavior, they help to understand the geometry of the high-dimensional energy landscapes. Earlier results were pioneered in the beautiful work of Auffinger-Ben Arous-Černý [5] in the spherical pure pp-spin model, whose Hamiltonian is defined as

HNG(σ)=1N(p−1)/2∑I∈[N]pGIσIH_N^G(\sigma)=\frac{1}{N^{(p-1)/2}}\sum_{I\in[N]^p}G_I\sigma_I

for σ∈NSN−1={σ∈RN:∥σ∥22=N}\sigma\in\sqrt{N}S^{N-1}=\{\sigma\in\mathbb{R}^N:\|\sigma\|_2^2=N\}, where σI:=∏α=1pσiα\sigma_I:=\prod_{\alpha=1}^{p}\sigma_{i_\alpha} for I=(i1,…,ip)I=(i_1,\ldots,i_p). Denote by ∇S\nabla_S and ∇S2\nabla_S^2 the spherical gradient and Hessian, respectively. Let

Crt⁡G((−∞,u)):={σ∈NSN−1:∇SHNG(σ)=0, HNG(σ)<Nu}.\operatorname{Crt}_G((-\infty,u)):=\{\sigma\in\sqrt{N}S^{N-1}:\nabla_SH_N^G(\sigma)=0,\ H_N^G(\sigma)<Nu\}.

be the collection of all critical points below the energy level u∈Ru\in\mathbb{R}. In [5 Theorem 2.8], it was established an explicit formula for the logarithmic asymptotics of the expected number of critical point, i.e.,

lim⁡N→∞1Nlog⁡E[Crt⁡G((−∞,u))]={12log⁡(p−1)+12+Ω(2uE∞)−u22,u≤−E∞,12log⁡(p−1)−p−24(p−1)u2,−E∞≤u≤0,12log⁡(p−1),u≥0.(1)\lim_{N\to\infty}\frac{1}{N}\log\mathbb{E}\left[\operatorname{Crt}_G((-\infty,u))\right]= \begin{cases} \frac{1}{2}\log(p-1)+\frac{1}{2}+\Omega\left(\frac{2u}{E_\infty}\right)-\frac{u^2}{2}, & u\le-E_\infty,\\ \frac{1}{2}\log(p-1)-\frac{p-2}{4(p-1)}u^2, & -E_\infty\le u\le0,\\ \frac{1}{2}\log(p-1), & u\ge0. \end{cases} \tag*{(1)}

where

E∞=2p−1p,μsc(dλ)=4−λ22π1{∣λ∣≤2} dλ(2)E_{\infty}=2\sqrt{\frac{p-1}{p}},\qquad\mu_{\mathrm{sc}}(d\lambda)=\frac{\sqrt{4-\lambda^{2}}}{2\pi}\mathbf{1}_{\{|\lambda|\leq2\}}\,d\lambda \tag*{(2)}

and Ω(x)\Omega(x) is the log-potential of semicircular law given by

Ω(x)=∫Rlog⁡∣λ−x∣ μsc(dλ)={x24−12−∣x∣4x2−4+log⁡(∣x∣+x2−42),∣x∣≥2,x24−12,∣x∣≤2.(3)\Omega(x)=\int_{\mathbb{R}}\log|\lambda-x|\,\mu_{\mathrm{sc}}(d\lambda)= \begin{cases} \frac{x^{2}}{4}-\frac{1}{2}-\frac{|x|}{4}\sqrt{x^{2}-4}+\log\left(\frac{|x|+\sqrt{x^{2}-4}}{2}\right), & |x|\geq2,\\ \frac{x^{2}}{4}-\frac{1}{2}, & |x|\leq2. \end{cases} \tag*{(3)}

In addition to this, the annealed complexity of saddle points was obtained in the same work [5] and later generalized to the mixed pp-spin model by Auffinger-Ben Arous [4]. Based on these, the study of complexity has received growing attention and many results have been made available. First, the second moment calculation of Crt⁡G((−∞,u))\operatorname{Crt}_{G}((-\infty,u)) was carried out in the pure pp-spin model by Subag [20], which concluded that the quenched and annealed complexities of the critical points indeed match with each other. More sophisticated formulas for the annealed complexity were also obtained in some Gaussian fields with isotropic increments by Auffinger-Zeng [6] as well as with non-rotationally invariant potential by Chen-Lu-Sen [12].

While the aforementioned results heavily rely on the Gaussianity of the Hamiltonians, it is a natural question to ask if the annealed complexity (1) will remain the same or not when GIG_I's are not Gaussian. Let {JI}I∈[N]p,N∈N\{J_I\}_{I\in[N]^p,N\in\mathbb{N}} be a collection of independent random variables independent of GIG_I's. By switching GG with JJ, let HNJH_N^J be defined in the same manner as HNGH_N^G. A reasonable setting to investigate the universality of the annealed complexity is to consider the pure pp-spin Hamiltonian associated to the interpolated disorder between GIG_I and JIJ_I,

Ht(σ):=1N(p−1)/2∑I∈[N]p(1−tGI+tJI)σI(4)H_t(\sigma):=\frac{1}{N^{(p-1)/2}}\sum_{I\in[N]^p}\left(\sqrt{1-t}G_I+\sqrt{t}J_I\right)\sigma_I \tag*{(4)}

for 0≤t≤10\leq t\leq1 and investigate the corresponding number of critical points,

Crt⁡t(B):=∣{σ∈NSN−1∣Ht(σ)∈NB,∇SHt(σ)=0}∣(5)\operatorname{Crt}_t(B):=\left|\left\{\sigma\in\sqrt{N}\mathbb{S}^{N-1}\mid H_t(\sigma)\in NB,\nabla_{\mathbb{S}}H_t(\sigma)=0\right\}\right| \tag*{(5)}

for all Borel set B⊆RB\subseteq\mathbb{R}. In this paper we show that the annealed complexity does exhibit universality for all 0<t<10<t<1 if JIJ_I's are strictly subgaussian. Here we recall that a random variable XX (of unit variance) is κSG2\kappa_{\mathrm{SG}}^{2}-subgaussian if

E[eλX]≤eκSG2λ2/2,∀λ∈R.\mathbb{E}\left[e^{\lambda X}\right]\leq e^{\kappa_{\mathrm{SG}}^{2}\lambda^{2}/2},\qquad\forall\lambda\in\mathbb{R}.

It is strict subgaussian if

κSG=1.\kappa_{\mathrm{SG}}=1.

The following is our main result.

Theorem 1.1. Let p≥3p\geq3. Let {JI}I∈[N]p,N∈N\{J_I\}_{I\in[N]^p,N\in\mathbb{N}} be independent with zero mean and unit variance. If they satisfy (SSG), then for any 0≤t<10\leq t<1.

lim⁡N→∞1Nlog⁡E[Crt⁡t((−∞,u))]=lim⁡N→∞1Nlog⁡E[Crt⁡0((−∞,u))].(6)\lim_{N\to\infty}\frac{1}{N}\log\mathbb{E}\left[\operatorname{Crt}_t((-\infty,u))\right]=\lim_{N\to\infty}\frac{1}{N}\log\mathbb{E}\left[\operatorname{Crt}_0((-\infty,u))\right]. \tag*{(6)}

Example 1.2. Assume that JIJ_I's are i.i.d. symmetric Racemacher, i.e., P(JI=±1)=1/2\mathbb{P}(J_I=\pm1)=1/2. It is centered with unit variance. Moreover, its moment generating function satisfies (SSG) since E[eλJ]=cosh⁡(λ)≤eλ2/2\mathbb{E}[e^{\lambda J}]=\cosh(\lambda)\leq e^{\lambda^{2}/2}. Then the conclusion of Theorem 1.1 holds.

Example 1.3. Assume that JIJ_I’s are i.i.d. sampled uniformly from (−3,3)(-\sqrt{3},\sqrt{3}). It is centered with unit variance. Moreover, its moment generating function satisfies (SSG) since E[eλJ]=sinh⁡(3λ)/(3λ)≤eλ2/2\mathbb{E}[e^{\lambda J}] = \sinh(\sqrt{3}\lambda)/(\sqrt{3}\lambda) \le e^{\lambda^2/2}. Thus, (6) holds.

References

  1. [1]Adler, R. J., and Taylor, J. E. Random Fields and Geometry. Springer Monographs in Mathematics. Springer, 2007.DOI
  2. [2]Aizenman, M., Peled, R., Schenker, J., Shamis, M., and Sodin, S. Matrix regularizing effects of Gaussian perturbations. Communications in Contemporary Mathematics 19, 3 (2017), 1750028.DOI
  3. [3]Anderson, G. W., Guionnet, A., and Zeitouni, O. An Introduction to Random Matrices, vol. 118 of Cambridge Studies in Advanced Mathematics. Cambridge University Press, 2010.DOI
  4. [4]Auffinger, A., and Ben Arous, G. Complexity of random smooth functions on the high-dimensional sphere. The Annals of Probability 41, 6 (2013), 4214–4247.DOI
  5. [5]Auffinger, A., Ben Arous, G., and Černý, J. Random matrices and complexity of spin glasses. Communications on Pure and Applied Mathematics 66, 2 (2013), 165–201.arxiv.org/abs/1003.1129
  6. [6]Auffinger, A., and Zeng, Q. Complexity of Gaussian random fields with isotropic increments. Communications in Mathematical Physics 402, 1 (2023), 951–993.arxiv.org/abs/2007.07668
  7. [7]Augeri, F., Guionnet, A., and Husson, J. Large deviations for the largest eigenvalue of sub-Gaussian matrices. Communications in Mathematical Physics 383, 2 (2021), 997–1050.arxiv.org/abs/1911.10591
  8. [8]Ben Arous, G., Bourgade, P., and McKenna, B. Exponential growth of random determinants beyond invariance. Probability and Mathematical Physics 3, 4 (2022), 731–789.DOI
  9. [9]Ben Arous, G., Dembo, A., and Guionnet, A. Aging of spherical spin glasses. Probability Theory and Related Fields 120, 1 (2001), 1–67.
  10. [10]Boucheron, S., Lugosi, G., and Massart, P. Concentration Inequalities: A Nonasymptotic Theory of Independence. Oxford University Press, Oxford, 2013.
  11. [11]Chen, W.-K., Dafnis, N., and Paouris, G. Improved Hölder and reverse Hölder inequalities for Gaussian random vectors. Advances in Mathematics 280 (2015), 643–689.DOI
  12. [12]Chen, W.-K., Lu, T.-L., and Sen, A. Complexity of the p-spin Hamiltonian with a non-rotationally invariant potential. arXiv preprint arXiv:2602.11036 (2026).arxiv.org/abs/2602.11036
  13. [13]Cook, N. A., Ducatez, R., and Guionnet, A. Full large deviation principles for the largest eigenvalue of sub-gaussian wigner matrices. arXiv preprint arXiv:2302.14823 (2023).arxiv.org/abs/2302.14823
  14. [14]Dembo, A., and Zeitouni, O. Large Deviations Techniques and Applications. Springer Science & Business Media, 2010.
  15. [15]Guionnet, A., and Husson, J. Large deviations for the largest eigenvalue of Rademacher matrices. The Annals of Probability 48, 3 (2020), 1436–1465.arxiv.org/abs/1810.01188
  16. [16]Guionnet, A., and Zeitouni, O. Concentration of the spectral measure for large matrices. Electronic Communications in Probability 5 (2000), 119–136.DOI
  17. [17]Ho, F.-H. Hessian descent for spherical spin glasses with uniform log-Sobolev disorder. arXiv preprint arXiv:2504.10625 (2025).arxiv.org/abs/2504.10625
  18. [18]Kim, T. Regularized bulk universality versus bounded-disorder nonuniversality for annealed complexity of spherical p-spin landscapes. arXiv preprint arXiv:2607.27613 (2026).arxiv.org/abs/2607.27613
  19. [19]Kivimae, P., and McKenna, B. A universal lower bound for the complexity of Gaussian-divisible spin glasses. To be posted on arXiv (2026).arxiv.org/abs/2610.23823
  20. [20]Subag, E. The complexity of spherical p-spin models—a second moment approach. The Annals of Probability 45, 5 (2017), 3385–3450.arxiv.org/abs/1504.02251
  21. [21]Vershynin, R. High-Dimensional Probability: An Introduction with Applications in Data Science, 2nd ed. Cambridge Series in Statistical and Probabilistic Mathematics. Cambridge University Press, 2026.

Paper details

vlnc:W_gdu8r2jw
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 7

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

ChatGPT (OpenAI)

View statement
Looking for related papers…
Contents