On the discrepancy principle for iterative and parametric methods to solve linear ill-posed equations (Q2564478)

The author examines a class of iteration and parametric methods for the linear ill-posed problem \(Au=f\) in a Banach space. This class contains the Richardson iteration \(u_{n+1} = u_n- \mu(Au_n-f)\), the implicit method \((I+ \mu A) u_{n+1} =u_n+ \mu f\), and others. The author shows that, under some conditions on the resolvent of \(A\), for given noise level of the right-hand term, the discrepancy principle defines a regularization method for the (possibly ill-posed) problem \(Au =f\). Thus the author extends to the case of Banach spaces results earlier known in the case of Hilbert spaces.