This paper is focuses on the computation of the positive moments of one-sidecorrelated random Gram matrices. Closed-form expressions for the moments can beobtained easily, but numerical evaluation thereof is prone to numericalstability, especially in high-dimensional settings. This letter provides anumerically stable method that efficiently computes the positive moments inclosed-form. The developed expressions are more accurate and can lead to higheraccuracy levels when fed to moment based-approaches. As an application, we showhow the obtained moments can be used to approximate the marginal distributionof the eigenvalues of random Gram matrices.
展开▼
