1
$\begingroup$

If we have a $P\times N$ matrix $\mathbf{A}$ whose elements $A_i$ are samples (in this case, P samples) from a multivariate gaussian distribution in an $N$ dimensional space, we can define the Wishart random symmetric matrix through

$$\mathbf{W}=\frac{1}{P}\mathbf{A}\mathbf{A}^\top.$$

Given this matrix, what is a good way of obtaining an order of magnitude estimation for the largest eigenvalue of $\mathbf{W}$ in the case where the number of samples $P$ is less than the degrees of freedom $N$ (i.e. $P<N$)?

In the end I'm looking to obtain an estimation for the order of magnitude of the eigenvalues $O(\rho(\lambda))(M,N)$ as a function of the number os samples M and the dimensionality of the space N.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.