Interval iterative methods for computing Moore-Penrose inverse (Q864808)

An interval iterative methods for computing the Moore-Penrose inverse of a full row (or column) rank matrix is presented. By modifying the classical Newton iteration by an interval method, the authors achieve higher numerical accuracy but the computation time is longer. They prove the convergence of the interval iteration. Computational results based on two randomly generated matrices (\(50 \times 60\) and \(500 \times 600\)) are given to exhibit the relatively better accuracy of the proposed algorithm.