Visually inspecting the search behavior of Harmony Search and its variants with Viz3D
Marcelo Lotif, André L. V. Coelho · 2011
The aim of this study is to assess the Harmony Search (HS) meta-heuristic and some of its variants when submitted to benchmark continuous optimization problems to reveal whether and how such variants change the patterns of search behavior exhibited by the canonical version. For this purpose, a new Visual Mining tool based on the Viz3D algorithm was developed to aid in the visualization of how the HS algorithms effectively explore the search space. The results achieved provide evidence that the gains in performance usually promoted by the HS variants are indeed related to noticeable modifications in the search behavior as displayed by the original version.