Parameter choice by discrepancy principles for the approximate solution of Ill-posed problems

A general a posteriori strategy for choosing the regularization parameter as a function of the error level is given which provides nearly the optimal rate of convergence.