An iterative method with truncation for infinite linear systems (Q1113611)

The authors consider an infinite system \(Tx=v\), where \(T=(t_{ij})\) is an infinite matrix and \(v\in l_{\infty}\). It is first established that if \(t_{ii}\), \(i=1,2,3,..\). are uniformly bounded away from zero and T is uniformly strictly diagonally dominant then T has a unique solution x. The system can be rewritten as \(x=Ax+b\), where \(A=D^{-1}F,\quad b=D^{-1}v,\quad D=diag(T),\) \(F=T-D\). Iterations of the form \(x^{(p+1)}=Ax^{(p)}+b,\) \(x^{(0)}=b\) and their truncations \(x^{(p+1,n)}=A_{(n)}x^{(p,n)}+b,\quad x^{(0,n)}=b\) are considered, where \(A_{(n)}\) is a truncation of A such that the upper left hand \(n\times n\) corner is left in place and zeroes are placed elsewhere. With two additional properties imposed on T upper bounds for \(\| x^{(p)}-x^{(p,n)}\|\) and \(\| x-x^{(p,n)}\|\) are established and convergence \(x^{(p,n)}\to x\) is proved. Two examples are given illustrating the choice of p and n for a desired accuracy.