Phase transition of the largest eigenvalue for nonnull complex sample covariance matrices
By Jinho Baik, Gérard Ben Arous, S. P'ech'e. 2004
1,186 citations recorded by the source corpus.
AbstractWe compute the limiting distributions of the largest eigenvalue of a complex Gaussian samplecovariance matrix when both the number of samples and the number of variables in each samplebecome large. When all but finitely many, say r, eigenvalues of the covariance matrix arethe same, the dependence of the limiting distribution of the largest eigenvalue of the samplecovariance matrix on those distinguished r eigenvalues of the covariance matrix is completelycharacterized in terms of an infinite sequence of new distribution functions that generalizethe Tracy-Widom distributions of the random matrix theory. Especially a phase transitionphenomena is observed. Our results also apply to a last passage percolation model and aqueuing model. 1 Introduction Consider M independent, identically distributed samples y 1 ,...,~y M , all of which are N ×1 columnvectors. We further assume that the sample vectors ~y k are Gaussian with mean µ and covarianceΣ, where Σ is a fixed N ×N positive matrix; the density of a sample ~y isp(~y) =1(2π)
Subfields: Discrete Mathematics and Combinatorics, Mathematical Physics, Statistics and Probability, Probabilistic & Bayesian Methods
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Discrete Mathematics and Combinatorics
In Mathematical Physics
In Statistics and Probability
In Probabilistic & Bayesian Methods
- Noureddine El Karoui
- Jun Yin
- George Kapetanios
- A. Onatski
- Boaz Nadler
- Jianfeng Yao
- Debashis Paul
- Tommaso Cai
Citations in the corpus
- The Full Spectrum of Deep Net Hessians At Scale: Dynamics with Sample Size (2018, background): “Much is known about its properties [2, 3, 13] and it is therefore of interest to know whether it indeed models the spectrum of the Hessian faithfully.”
- Empirical Analysis of the Hessian of Over-Parametrized Neural Networks (2017, background): “One of the earliest relaxations appear in [Baik et al., 2005].”
- Outlier Eigenvalues for Deformed I.I.D. Random Matrices (2014, neutral): “Note that motivated by the seminal article of Baik, Ben Arous, and Péché [6], previous works are devoted to the study of the outlier eigenvalues of deformed random Hermitian matrix models; see notably [4, 7–11, 18, 20–22, 28, 34–37, 42, 44, 46, 47].”
- Cleaning large correlation matrices: tools from random matrix theory (2016, background): “It is interesting to compare this with the well-known “Baik-Ben Arous-Péché” (BBP) result on large outliers [118], which reads (see Eq.”
- Estimating structured high-dimensional covariance and precision matrices: Optimal rates and adaptive estimation (2016, background): “But unlike estimation problems there are only few optimality results on testing, see Baik et al. (2005), El Karoui (2007), Cai and Ma (2013) and Onatski et al. (2013).”
- PCA Consistency for Non-Gaussian Data in High Dimension, Low Sample Size Context (2009, background): “…substantial work has been done on the asymptotic behavior of eigenvalues of the sample covariance matrix in the limit as d → ; see Johnstone (2001), Baik et al. (2005), and Paul (2007) for Gaussian assumptions, and Baik and Silverstein (2006) for non-Gaussian but i.i.d. assumptions when d and n…”
- Limits of spiked random matrices I (2010, background): “10 (BBP [9], Tracy and Widom [53,54], Wang [56]) We have”
- The Dynamics of Learning: A Random Matrix Approach (2018, background): “One may thus expect to have a smooth and reliable behavior of the well-trained network for c close (3)This phenomenon has been largely observed in random matrix theory and is referred to as “phase transition”(Baik et al., 2005).”
- On finite rank deformations of Wigner matrices (2011, methodology): “ i.i.d. entries, in particular [30], [19], [26], [12], [13], [14], [8], [9], [10], [35]. We also note several papers on the eigenvalues of sample covariance matrices of spiked population models ([3], [5], [6], [29]). This manuscript can be viewed as a companion paper to our recent works [31] and [28”
- A pedestrian's view on interacting particle systems, KPZ universality and random matrices (2008, background): “A proof of the Prähofer–Spohn conjecture for the TASEP was recently presented by Ben Arous and Corwin [4] (see also [5, 48, 96] for earlier partial results).”
- Spiked separable covariance matrices and principal components (2019, background): “s spiked covariance matrix model [26]. In this model, a few spikes, i.e., eigenvalues detached from the bulk eigenvalue spectrum, are added to Ar. Since the seminal work of Baik, Ben Arous and P´ech´e[4], it is now well-understood that the extremal eigenvalues undergo a so-called BBP transition alon”
- Community detection and stochastic block models: recent developments (2017, result): “• Note that Theorem 53 requires merely diverging degrees (arbitrarily slowly), in contrast to results from random matrix theory such as Baik et al. (2005) that would require poly-logarithmic degrees to extract communities from the spiked Wigner model.”
- Painlevé formulas of the limiting distributions for nonnull complex sample covariance matrices (2005, background): “The definition (1.11) involves the functions s(m) and t(m) and it is not transparent that the formula is symmetric in w1, . . . , wk, which should be the case from its origin in the sample covariance matrix [1] (see also Section 2 below).”
- Singular vector and singular subspace distribution for the matrix denoising model (2018, background): “Since the seminal work of Baik, Ben Arous and Péché [4], it is now well-understood that the extreme eigenvalues undergo a so-called BBP transition along with the change of the strength of the deformation.”
- Scaling Limit for the Space-Time Covariance of the Stationary Totally Asymmetric Simple Exclusion Process (2005, background): “ition 3.2. By Proposition C.1 Km,d is a similarity transform of the Laguerre kernel. Therefore Km,d = (Km,d)2, kKm,dk = 1, PuKm,dPu is trace class, and all eigenvalues of PuKm,dPu are in the interval [0,1]. Thus we only prove that, if u>0, 1 is not in the spectrum of PuKm,dPu, which is accomplish”
- High Dimensional Statistical Inference and Random Matrices (2006, background): “[7] describe, for complex data, a ‘phase transition’ that occurs between these two extremes.”