Parameter identification with weightless regularization (Q2761940)

The author wants to avoid the classical technique for solving a parameter identification problem. Usually the problem is converted into an objective function, which has to be minimized using an optimization method. In this approach difficulties arise if measurement data and/or the model contain large errors.NEWLINENEWLINENEWLINETo overcome this problem the paper proposes to add a regularization term that consists of a function multiplied with weighting factors such as a regularization parameter, to the objective function. This term makes the functional smooth so that a classical optimization method can obtain an appropriate parameter set in a more stable fashion. Numerical examples show the efficiency of this approach.