Skip to Main Content
Table 1:
Summary of the contributions of this study with respect to existing concepts.
ConceptAvailable formThis study
Taboo regions (Glover, 1986) (Siarry and Berthiau, 1997) (Ahrari et al., 2009) The size of the taboo regions is fixed or decreases dynamically. Taboo regions are adaptively resized and reshaped. 
 Taboo regions are similar for all solutions. Taboo regions differ for different subpopulations. 
 The size of the taboo regions is independent of the basin size. The size of the taboo region is adapted to the basin size. 
 Taboo regions are the recently visited regions. Taboo regions are the previously identified basins and the center of fitter subpopulations. 
 All taboo regions are checked. Only critical taboo regions are checked. 
Core algorithm (CMSA-ES) (Beyer and Sendhoff, 2008) Comma selection is employed. Elite selection is employed. 
 The global step size is adapted by simple averaging. A bias compensation is applied in adaptation of the global step size. 
Restarts (Auger and Hansen, 2005) Independent restarts are performed and population size increases. Outcome of a restart affects the subsequent restarts. (Similarly employed) 
Hill-valley function (Ursem, 1999) Checks whether two solutions share the same basin. (Similarly employed) 
Semi-random initialization (Ahrari et al., 2009) Initialization maximizes the initial distance to other subpopulations. Initialization maximizes the initial distance to other subpopulations and the archived points. 
ConceptAvailable formThis study
Taboo regions (Glover, 1986) (Siarry and Berthiau, 1997) (Ahrari et al., 2009) The size of the taboo regions is fixed or decreases dynamically. Taboo regions are adaptively resized and reshaped. 
 Taboo regions are similar for all solutions. Taboo regions differ for different subpopulations. 
 The size of the taboo regions is independent of the basin size. The size of the taboo region is adapted to the basin size. 
 Taboo regions are the recently visited regions. Taboo regions are the previously identified basins and the center of fitter subpopulations. 
 All taboo regions are checked. Only critical taboo regions are checked. 
Core algorithm (CMSA-ES) (Beyer and Sendhoff, 2008) Comma selection is employed. Elite selection is employed. 
 The global step size is adapted by simple averaging. A bias compensation is applied in adaptation of the global step size. 
Restarts (Auger and Hansen, 2005) Independent restarts are performed and population size increases. Outcome of a restart affects the subsequent restarts. (Similarly employed) 
Hill-valley function (Ursem, 1999) Checks whether two solutions share the same basin. (Similarly employed) 
Semi-random initialization (Ahrari et al., 2009) Initialization maximizes the initial distance to other subpopulations. Initialization maximizes the initial distance to other subpopulations and the archived points. 
Close Modal

or Create an Account

Close Modal
Close Modal