Exponential growth of random determinants beyond invariance

Abstract

We give simple criteria to identify the exponential order of magnitude of the absolute value of the determinant for wide classes of random matrix models, not requiring the assumption of invariance. These include Gaussian matrices with covariance profiles, Wigner matrices and covariance matrices with subexponential tails, Erdős-Rényi and $d$-regular graphs for any polynomial sparsity parameter, and non-mean-field random matrix models, such as random band matrices for any polynomial bandwidth. The proof builds on recent tools, including the theory of the Matrix Dyson Equation as developed in [Ajanki, Erdős, Krüger 2019]. We use these asymptotics as an important input to identify the complexity of classes of Gaussian random landscapes in our companion papers [Ben Arous, Bourgade, McKenna 2021; McKenna 2021].

arXiv commentApplication of main result to Wigner matrices actually requires subexponential moments (but see new Appendix B for a related result under weaker assumptions). New results on $d$-regular random graphs. 48 pages

References in corpus (19)

• Spectral statistics of Erdős-Rényi graphs I: Local semicircle law
• Random incidence matrices: moments of the spectral density
• Random Matrices with Slow Correlation Decay
• Random matrices: Law of the determinant
• Law of Log Determinant of Sample Covariance Matrix and Optimal Estimation of Differential Entropy for High-Dimensional Gaussian Distributions
• Local semicircle law for random regular graphs
• Stability of the Matrix Dyson Equation and Random Matrices with Correlations
• Spectrum of non-Hermitian heavy tailed random matrices
• Free energy and complexity of spherical bipartite models
• A large deviation principle for Wigner matrices without Gaussian tails
• Optimal local law and central limit theorem for $\beta$-ensembles
• Asymptotic behavior of random determinants in the Laguerre, Gram and Jacobi ensembles
• Edge rigidity and universality of random regular graphs of intermediate degree
• The Loss Surfaces of Neural Networks with General Activation Functions
• The logarithmic law of random determinant
• Manifolds in high dimensional random landscape: complexity of stationary points and depinning
• The number of saddles of the spherical $p$-spin model
• A spin-glass model for the loss surfaces of generative adversarial networks
• Counting equilibria of large complex systems by instability index

Cited by in corpus (3)

• Concentration of the complexity of spherical pure $p$-spin models at arbitrary energies
• Complexity of Gaussian random fields with isotropic increments: critical points with given indices
• On-Site Potential Creates Complexity in Systems with Disordered Coupling