Effiziente Ableitungsbestimmung bei hochdimensionaler nichtlinearer Optimierung

Abstract

Calculating or approximating derivative information in an efficient way is one of the major tasks when high dimensional nonlinear optimization problems are to be solved. The thesis shows various new ways both in the area of finite differences and in the area of update formalae to tackle this problem. One of the results in this thesis is the development of a sparse BFGS-similar method for which superlinear convergence properties are shown.