Finding the Moore-Penrose inverse by a new matrix iteration (Q2354185)

This paper develops an iterative procedure to compute pseudo-inverses of matrices, square, nonsingular or singular, and rectangular. The method has convergence order 9 and uses 7 matrix times matrix multiplications in each iteration step, thereby achieving an information index of 9/7 that exceeds 1 for the first time when computing Moore-Penrose inverses. There are results on good start-up values for the iteration as well as termination criteria comparisons. Run-time comparisons with other schemes are included.