Descent heuristics for unconstrained minimization

Semidefinite relaxations often provide excellent starting points for nonconvex problems with multiple local minimizers. This work aims to find a local minimizer within a certain neighborhood of the starting point and with a small objective value. Several approaches are motivated and compared with each other.

Citation

Report, Mathematisches Institut, Universitaet Duesseldorf, August 2008.

Article

Download

View PDF