A self-correcting matrix iteration for the Moore-Penrose generalized inverse Original Research Article

Let Xk be the kth iterate of an algorithm for finding the Moore-Penrose generalized inverse of a singular matrix A. The popular algorithm Xk + 1 = Xk(2I − AXk) has a first-order error component, but the more complicated algorithm given here has no first-order error component for general A and Xk.