An eigenspace divide-and-conquer approach for large-scale optimization

Divide-and-conquer-based (DC-based) evolutionary algorithms (EAs) have achieved notable success in dealing with large-scale optimization problems (LSOPs). However, the appealing performance of this type generally requires a high-precision decomposition problem, which is still challenging task for existing methods. This study attempts to address above issue from different perspective and proposes an eigenspace divide-and-conquer (EDC) approach. Different DC-based that perform original solution space, EDC first establishes by conducting singular value on set high-quality solutions selected recent generations. Then it transforms problem into eigenspace, thus significantly weakens dependencies among corresponding eigenvariables. Accordingly, these eigenvariables can be efficiently grouped simple random strategy each resulting subproblems addressed more easily traditional EA. To verify efficiency EDC, comprehensive experimental studies were conducted two sets benchmark functions. Experimental results indicate robust its parameters has good scalability dimension. The comparison several state-of-the-art further confirms pretty competitive performs better complicated LSOPs.