Files
Abstract
Recently developed quasi-Newton algorithms for unconstrained optimization focus on the solution of badly scaled problems. A uniform numerical comparison of these algorithms is performed to get insight in their relative behaviour and to verify emperically their ability to solve badly scaled problems. A suitable battery of testproblems is suggested. The design of the experiments and their results are preceded by a short description of the theoretical backgrounds as they were developed by Oren and Luenberger. Recent alternatives of Oren and Spedicato and of Shanno and Phua are mentioned The classical DFP- and BFGS-algorithms are considered as well.