We demonstrate that eigenvalues minimize information entropy of the main diagonal elements of a symmetric matrix over all possible orthogonal transforms. This legitimizes the use of the exponential entropy of standardized eigenvalues as a measure of effective covariance matrix dimensionality. In...