The order of convergence for Landweber scheme with \(\alpha,\beta\)-rule (Q2428065)

Many problems in image reconstruction can be represented by a system of linear algebraic equations \( Ax = b. \) The authors consider the Landweber scheme for solving such a system, \[ x^{m+1} = x^{m} + \lambda_m V^{-1} A^* W (b-Ax^{m}), \] where \(\lambda_m\) is a relaxation coefficient and \(V\) and \(W\) are given Hermitian positive definite matrices. Using different choices of \(V\) and \(W\), the Landweber scheme leads to some well-known iterative methods (e.g., Cimmino's method). The authors study the Landweber scheme with variable relaxation coefficients~\(\lambda_m\). To stop the iterations, they use a stopping criterion called \(\alpha, \beta\)-rule. The order of convergence of such a scheme is investigated. The authors conclude that the \(\alpha, \beta\)-rule is the optimal stopping criterion.