Speaker:
Sandrine Peche (Institut Fourier, Grenoble)
Title:
Non white sample covariance matrices
Abstract:
Consider a large sample covariance matrix $M_N= \Sigma^{1/2}X(\Sigma^{1/2}X)^*/N$, where $X$ is a $N \times p$ random matrix with i.i.d. normalized entries and $\Sigma$ is a $N \times N$ positive definite matrix. We will study some impact of the population covariance $\Sigma$ on the largest eigenvalues of $M_N$ as well as on its eigenvectors. This is recent work with D. F\'eral and O. Ledoit.