Diagonalize the rate matrix, K, from either the matrix K or the symmetric rate matrix, S.
If which == ‘K’, the first argument should be the rate matrix, K, and pi is ignored. If which == ‘S’, the first argument should be the symmetric rate matrix, S. This can be build using buildK(... which=’S’), and pi should contain the equilibrium distribution (left eigenvector of K with eigenvalue 0, and also the last n elements of exptheta).
Whichever is supplied the return value is the eigen decomposition of K. The eigendecomposition of S is not returned.
Using the symmetric rate matrix, S, is somewhat faster and more numerically stable.
| Returns: | w : array
U : array, size=(n,n)
V : array, size=(n,n)
|
|---|