A unified optimization framework for population-based methods

Jin Sun, Qianchuan Zhao, Peter B. Luh

Abstract: Combinatorial optimization problems arise in many applications such as task assignment, facility location, and elevator scheduling. A wide variety of population-based solution methods have been developed, either instance-based (e.g., genetic algorithm (GA) and particle swarm optimization (PSO)) or model-based (e.g., ant colony optimization (ACO) and estimation of distribution algorithms (EDAs)). Their various mechanisms make it difficult to analyze and compare these methods and to extend the advancement in one method to another. To this end, a unified optimization framework towards representing these seemingly different methods is established as iteratively sampling and updating of a population distribution. This framework is then innovatively instantiated with PSO from the instance-based category and EDA from the model-based category. Finally, the possible use and the finite time performance analysis of the unified framework are discussed.