Models of parallel chaotic iteration methods (Q1103326)

Linear systems \(Ax=b\) where the matrix is monotonic i.e. has a nonnegative inverse are considered. The convergence of parallel algorithms, where each processor performs several iterations updating a part of x between each global update of the whole of x, is studied and proved in two cases. Both the free steering method of \textit{A. M. Ostrowski} [J. Math. Anal. Appl. 2, 351-369 (1961; Zbl 0100.333)], and the chaotic relaxation by \textit{D. Chazan} and \textit{W. Miranker} [Linear Algebra Appl. 2, 199-222 (1969; Zbl 0225.65043)] as well as the block Jacobi iteration are covered by the theory.