On the choice of the regularization parameter for iterated Tikhonov regularization of ill-posed problems (Q1085952)

The iterated Tikhonov method for the solution of the ill-posed problem \(Tx=y\), where T is a bounded linear operator from one Hilbert space X to another, Y, consists in solving \((T^*T+\alpha I)x^ j_{\alpha,\delta}=T^*y_{\delta}+\alpha x^{j-1}_{\alpha,\delta}\) for \(x^ j_{\alpha,\delta}\), where \(1\leq j\leq m\in {\mathbb{N}}\), \(\alpha,\delta >0,\) \(x^ o_{\alpha,\delta}:=0\) and \(\| y_{\delta}-y\| \leq \delta\). The author proposes a method for choosing the regularization parameter \(\alpha\) in such a way that the optimal convergence rate \(\| x^ m_{\alpha,\delta}-x\| =O(\delta^{2n/(2n+1)})\) is achieved for \(x\in R((T^*T)^ n)\), where \(0<n\leq m\) and x is the least squares minimum norm solution of \(Tx=y\).