-
Another Marčenko-Pastur law for Kendall's tau
Authors:
Pierre Bousseyroux,
Tomas Espana,
Matteo Smerlak
Abstract:
Bandeira et al. (2017) show that the eigenvalues of the Kendall correlation matrix of $n$ i.i.d. random vectors in $\mathbb{R}^p$ are asymptotically distributed like $1/3 + (2/3)Y_q$, where $Y_q$ has a Marčenko-Pastur law with parameter $q=\lim(p/n)$ if $p, n\to\infty$ proportionately to one another. Here we show that another Marčenko-Pastur law emerges in the "ultra-high dimensional" scaling limi…
▽ More
Bandeira et al. (2017) show that the eigenvalues of the Kendall correlation matrix of $n$ i.i.d. random vectors in $\mathbb{R}^p$ are asymptotically distributed like $1/3 + (2/3)Y_q$, where $Y_q$ has a Marčenko-Pastur law with parameter $q=\lim(p/n)$ if $p, n\to\infty$ proportionately to one another. Here we show that another Marčenko-Pastur law emerges in the "ultra-high dimensional" scaling limit where $p\sim q'\, n^2/2$ for some $q'>0$: in this quadratic scaling regime, Kendall correlation eigenvalues converge weakly almost surely to $(1/3)Y_{q'}$.
△ Less
Submitted 25 March, 2025; v1 submitted 24 March, 2025;
originally announced March 2025.
-
Distribution of the Diagonal Entries of the Resolvent of a Complex Ginibre Matrix
Authors:
Pierre Bousseyroux,
Jean-Philippe Bouchaud,
Marc Potters
Abstract:
The study of eigenvalue distributions in random matrix theory is often conducted by analyzing the resolvent matrix $ \mathbf{G}_{\mathbf{M}}^N(z) = (z \mathbf{1} - \mathbf{M})^{-1} $. The normalized trace of the resolvent, known as the Stieltjes transform $ \mathfrak{g}_{\mathbf{M}}^N(z) $, converges to a limit $ \mathfrak{g}_{\mathbf{M}}(z) $ as the matrix dimension $ N $ grows, which provides th…
▽ More
The study of eigenvalue distributions in random matrix theory is often conducted by analyzing the resolvent matrix $ \mathbf{G}_{\mathbf{M}}^N(z) = (z \mathbf{1} - \mathbf{M})^{-1} $. The normalized trace of the resolvent, known as the Stieltjes transform $ \mathfrak{g}_{\mathbf{M}}^N(z) $, converges to a limit $ \mathfrak{g}_{\mathbf{M}}(z) $ as the matrix dimension $ N $ grows, which provides the eigenvalue density $ ρ_{\mathbf{M}} $ in the large-$ N $ limit. In the Hermitian case, the distribution of $ \mathfrak{g}_{\mathbf{M}}^N(z) $, now regarded as a random variable, is explicitly known when $ z $ lies within the limiting spectrum, and it coincides with the distribution of any diagonal entry of $ \mathbf{G}_{\mathbf{M}}^N(z) $.
In this paper, we investigate what becomes of these results when $ \mathbf{M} $ is non-Hermitian. Our main result is the exact computation of the diagonal elements of $ \mathbf{G}_{\mathbf{M}}^N(z) $ when $ \mathbf{M} $ is a Ginibre matrix of size $ N $, as well as the high-dimensional limit for different regimes of $ z $, revealing a tail behavior connected to the statistics of the left and right eigenvectors. Interestingly, the limit distribution is stable under inversion, a property previously observed in the symmetric case. We then propose two general conjectures regarding the distribution of the diagonal elements of the resolvent and its normalized trace in the non-Hermitian case, both of which reveal a symmetry under inversion.
△ Less
Submitted 28 November, 2024;
originally announced November 2024.
-
Spectral Initialization for High-Dimensional Phase Retrieval with Biased Spatial Directions
Authors:
Pierre Bousseyroux,
Marc Potters
Abstract:
We explore a spectral initialization method that plays a central role in contemporary research on signal estimation in nonconvex scenarios. In a noiseless phase retrieval framework, we precisely analyze the method's performance in the high-dimensional limit when sensing vectors follow a multivariate Gaussian distribution for two rotationally invariant models of the covariance matrix C. In the firs…
▽ More
We explore a spectral initialization method that plays a central role in contemporary research on signal estimation in nonconvex scenarios. In a noiseless phase retrieval framework, we precisely analyze the method's performance in the high-dimensional limit when sensing vectors follow a multivariate Gaussian distribution for two rotationally invariant models of the covariance matrix C. In the first model C is a projector on a lower dimensional space while in the second it is a Wishart matrix. Our analytical results extend the well-established case when C is the identity matrix. Our examination shows that the introduction of biased spatial directions leads to a substantial improvement in the spectral method's effectiveness, particularly when the number of measurements is less than the signal's dimension. This extension also consistently reveals a phase transition phenomenon dependent on the ratio between sample size and signal dimension. Surprisingly, both of these models share the same threshold value.
△ Less
Submitted 22 March, 2024;
originally announced March 2024.
-
Asymptotics of non-integer moments of the logarithmic derivative of characteristic polynomials over $SO(2N+1)$
Authors:
Emilia Alvarez,
Pierre Bousseyroux,
Nina C. Snaith
Abstract:
This work computes the asymptotics of the non-integer moments of the logarithmic derivative of characteristic polynomials of matrices from the $SO(2N+1)$ ensemble. It follows from work of Alvarez and Snaith who computed the asymptotics of the integer moments of the same statistic over both $SO(N)$ ensembles as well as the $USp(2N)$ ensemble.
This work computes the asymptotics of the non-integer moments of the logarithmic derivative of characteristic polynomials of matrices from the $SO(2N+1)$ ensemble. It follows from work of Alvarez and Snaith who computed the asymptotics of the integer moments of the same statistic over both $SO(N)$ ensembles as well as the $USp(2N)$ ensemble.
△ Less
Submitted 14 March, 2023;
originally announced March 2023.
