Escaping Local Optima with Local Search: A Theory-Driven Discussion

Local search is the most basic strategy in optimization settings when no specific problem knowledge employed. While this finds good solutions for certain problems, it generally suffers from getting stuck local optima. This stagnation can be avoided if modified. Depending on landscape, different modifications vary their success. We discuss several features of landscapes and give analyses as examples how they affect performance search. consider modifying random by restarting considering larger radii. The landscape we analyze include number optima, distance between well around a optimum. For each feature, show which handle them do not.