A convergence criterion for approximations in the residual method for linear ill-posed problems (Q4339998)

For a closed linear operator \( T:H \supset D(T) \to H \) with dense domain of definition \( D(T) \) in a Hilbert space \( H \), and non-closed range \( R(T) \), the subject under consideration is the equation \[ Tu = f \tag{1} \] which may be ill-posed in general. It is supposed that only an approximation \( f_\delta \in H \) for \( f \in R(T) \) is available that satisfies \( |f_\delta - f |< \delta \), where \( \delta > 0 \) is a known noise level, and \( R(T) \) denotes the range of \( T \). For the regularization of equation (1), the residual method \[ |u |\to \min, \quad u \in H, \qquad |T u - f_\delta |\leq \delta, \tag{2} \] as well as two approximation schemes for (2) are considered, \[ |u |\to \min, \qquad u \in H_n, \qquad |T_n u - f^n_\delta |\leq \delta \qquad (n =0, 1, \ldots), \tag{3} \] and \[ |u |\to \min, \qquad u \in H, \qquad |T_n u - f^n_\delta |\leq \delta \qquad (n =0, 1, \ldots). \tag{4} \] Here \( \delta > 0 \) is fixed, \( f_\delta^n \in H \) satisfies \( |f^n_\delta - f_\delta |\to 0 \) as \( n \to \infty \), \( T_n: H \to H \) is a linear bounded operator, and \( H_n \subset H \) is a finite-dimensional subspace of \( H \). The author provides sufficient and necessary conditions for the convergence of solutions of the minimization problems (3) and (4), respectively, to a solution of (2) as \( n \to \infty \).