Spectral measure of heavy tailed band and covariance random matrices (Q2391141)

The authors study the asymptotic behavior of the appropriately scaled and possibly perturbed spectral measure of large random real symmetric matrices with heavy tailed entries. In particular, they consider the \(N\)-by-\(N\) matrix \(Y\) with entries \(\sigma(i/N,j/N)x_{i,j}\), where \(\sigma:[0,1]\times[0,1]\rightarrow\mathbb R\) is a square integrable measurable symmetric function, \(x_{i,j}\) represents an infinite array of independent and identically distributed real variables. The authors show that for a random diagonal matrix \(D\) and with appropriate rescaling \(a\), the probability measure \(\mu_{a^{-1}Y+D}\) converges in mean towards a limiting probability measure, which is characterized. As a special case, they derive and analyze the almost sure limiting spectral density for empirical covariance matrices with heavy tailed entries.