Cooperative micro-differential evolution for high-dimensional problems
Konstantinos E. Parsopoulos · 2009
High–dimensional optimization problems appear very often in demanding applications. Although evolutionary algori-thms constitute a valuable tool for solving such problems, their standard variants exhibit deteriorating performance as dimension increases. In such cases, cooperative approaches have proved to be very useful, since they divide the com-putational burden to a number of cooperating subpopula-tions. In contrast, Micro–evolutionary approaches consti-tute light versions of the original evolutionary algorithms that employ very small populations of just a few individu-als to address optimization problems. Unfortunately, this property is usually accompanied by limited efficiency and proneness to get stuck in local minima. In the present work, an approach that combines the basic properties of coopera-tion and Micro-evolutionary algorithms is presented for the Differential Evolution algorithm. The proposed Cooperative Micro–Differential Evolution approach employs small coope-rative subpopulations to detect subcomponents of the origi-nal problem solution concurrently. The subcomponents are combined through cooperation of subpopulations to build complete solutions of the problem. The proposed approach is illustrated on high-dimensional instances of five widely used test problems with very promising results. Compari-sons with the standard Differential Evolution algorithm are also reported and their statistical significance is analyzed.