| Popis: |
A multistep method of least squares is described in [1]. It accelerates the convergence of iterative process $$ x_{n + 1} = Tx_n + f, n = 0,1,...$$ (1) of solving a system of linear algebraic equations (SLAE) Ax=b. Here, A, T are square matrices, b is a right-hand side parts vector, x 0 is an initial approximation for the exact solution x. In this method, correction y n * is added to a current approximation of x n every k steps, i.e. x n *=x n +y n *. Vector x n * is an improved approximation for the n-th iteration here. The correction is a linear combination of k residual vectors r i =Tx i +f−x i =x i+1−x i , where i=(n−k),…,(n−1), k |
