The computation of Drazin inverse and its application in Markov chains (Q864771)

The Drazin inverse \(A^D\) of the square matrix \(A\) is a matrix \(X\) satisfying \(A^kXA=A^k\), \(XAX=X\), and \(AX=XA\). \(k\) is called the index of \(A\), and when \(k=1\), \(A^D\) is equal to the group inverse \(A^\#\). In this paper, first a representation for the projector \(AA^D\) is obtained and this gives rise to two new algorithms for computing \(A^D\). One is based on an equivalence transformation of \(A^k\) where the transformation matrices are given by matrix series expansions. The second one is an iterative technique using hyperpowers. Convergence proofs for the methods are given. An efficient acceleration technique is given and numerical examples showing the effectiveness for the computation of Markov chains.