A discrete estimation of distribution particle swarm optimization for combinatorial optimization problems

The philosophy behind the original particle swarm optimization (PSO) is to learn from individual's own experience and the best individual experience in the whole swarm. Estimation of distribution algorithms (EDAs) generate new solutions from a probability model which characterizes the distribution of promising solutions in the search space at each generation. In this paper, a discrete estimation of distribution particle swarm optimization algorithm (DEDPSO) is proposed for combinatorial optimization problems. The proposed algorithm combines the statistical information collected from the local best solutions information of all individuals and the global best solution information found so far in the whole swarm. The results show that the proposed algorithm has superior performance to other discrete PSOs.