Heterogeneous Island Model with Re-planning of Methods

Abstract

We propose a heterogeneous island model where each of the islands can run a different optimization algorithm. The distributed computation is managed by a central planner, that re-plans the methods during the run of the algorithm -- less successful methods are removed while new instances of more successful methods are added.

We show that this re-planning improves the performance of the optimization algorithm.

Keywords

• Parallel evolutionary algorithms
• island model
• algorithm selection
• iterative re-planning
• adaptive system.