Techniques and Tools for Local Search Landscape Visualization and Analysis

Because of their high dimensionality, combinatorial optimization problems are often difficult to analyze, and the researcher's intuition is insufficient to grasp the relevant features. In this paper we present and discuss a set of techniques for the visualization of search landscapes aimed at supporting the researcher's intuition on the behavior of a Stochastic Local Search algorithm applied to a combinatorial optimization problem. We discuss scalability issues posed by the size of the problems and by the number of potential solutions, and propose approximate techniques to overcome them. Examples generated with an application (available for academic use) are presented to highlight the advantages of the proposed approach. © 2009 Springer Berlin Heidelberg.