A posteriori stopping rule for regularized fixed point iterations (Q2492434)

The authors study the iteratively regularized fixed-point scheme \(x_{n+1}=x_n-\alpha_n \{ F(x_n)-f_{\delta}+\varepsilon_n (x_n-x_0) \}\) for solving an ill-posed operator equation \(F(x)=f\) with a monotone and Lipschitz-continuous operator \(F: H \to H\) acting in a Hilbert space \(H\). Here, \(f_{\delta} \in H\) is an approximation to the true element \(f\). The iteration is combined with the modified discrepancy principle \(\| F(x_{N(\delta)})-f_{\delta} \|^2 \leq \tau \delta < \| F(x_n)-f_{\delta} \|^2\), \(0 \leq n <N(\delta)\), \(\tau>1\). It is shown that \(\lim_{\delta \to 0} \| x_{N(\delta)}-x^*\|=0\), where \(x^*\) is a solution to the original equation.