No label defined (Q4457828)

The author studies the constrained linear inverse problem \[ Ax=b,\;x\in K\subset H_1,\;b\in H_2, \] where \(K\) is a closed convex set, \(H_1,H_2\) are real Hilbert spaces and the linear bounded operator \(A:H_1\to H_2\) has non-closed range. Being this an ill-posed problem, Tikhonov regularization must be employed, namely \[ \min_{x\in K}\frac{1}{2}\| Ax-b\| ^2+\frac{1}{2}\alpha\| x\| ^2. \] This formulation leads to results of existence, uniqueness and continuous dependence on the data, as well as to: (i) convergence to the least-squares solution of the original problem; (ii) asymptotic estimates for noisy data. The article aim is to build fixed point approximations to the solution of the Tikhonov constrained regularized problem, for which a convergence-in-norm result is reached. Further, similar results are obtained for a generalized constrained Tikhonov regularization, namely \[ \min_{x\in K}\tfrac{1}{2}\| Ax-b\| ^2+ \alpha\theta(x), \] where \(\theta : H_1 \to\mathbb R^+\) is strictly convex, coercive, satisfies \(\theta(0)=0\), is differentiable, with the estimates \[ \langle x-y, \theta'(x)-\theta'(y)\rangle\geq\| x-y\| ^2,\qquad \| \theta'(x)-\theta'(y)\| \geq L\| x-y\|, \] holding for \(\eta, L >0\).