Semi-iterative methods for the approximate solution of ill-posed problems (Q1084837)

We study two-step iterative methods of the form \[ x_{n+1}=T^*T(\alpha x_ n+\gamma x_{n-1})+\beta x_ n(1- \beta)x_{n-1}-(\alpha +\gamma)T^*y \] for the approximate solution of ill-posed or ill-conditioned linear equations \(Tx=y\) in (infinite or finite dimensional) Hilbert spaces. We present results on convergence, convergence rates and on the influence of perturbed data. We introduce a general method for the comparison of different iterative methods: Let M and N be two iterative methods which generate for each right hand side y of the equation \(Tx=y\) sequences \((x^ M_ n)\) and \((x^ N_ n)\) which converge to the solution x of this equation. Let \((R_ n)\) be a sequence of linear operators which assign to each remainder \(x^ M_ n- x\) the remainder \(x^ N-x\) by \(x^ M_ n-x=R_ n(x^ N_ n-x)\). Then the method M is said to converge uniformly faster than the method N, if \(\lim \| R_ n\| =0\), and the method M is said to converge nonuniformly faster than the method N, if lim sup\(\| R_ n\| >0\), but lim \(R_ nv=0\) for each v. This principle is applied to different two-step iterative methods.