Linear shrinkage of the eigenvalues.

<p>Shrinkage of the eigenvalues of the MAP of the covariance matrix as <i>r</i> = <i>α</i>. When <i>n</i> < <i>p</i>, all eigenvalues increase when <i>r</i> is small but when <i>r</i> increases, the increment gets more moderate (black lines). When <i>n</i> increases, shrunken eigenvalues get closer to the sample eigenvalues and the growth of <i>r</i> does not significantly affect the estimated eigenvalues (red lines). When <i>n</i> ≫ <i>p</i>, estimated eigenvalues start to coincide to the sample eigenvalues (blue lines).