A divide-and-conquer approach for the computation of the Moore-Penrose inverses

摘要

In this paper, we present a divide-and-conquer method for computing the Moore-Penrose inverse of a bidiagonal matrix. Working together with the effective parallel algorithms for the reduction of a general matrix to the bidiagonal matrix, the proposed method provides a new parallel approach for the computation of the Moore-Penrose inverse of a general matrix. This new approach was implemented in the CUDA environment and a significant speedup was observed on randomly generated matrices.