Experimental Evaluation of ACO for Continuous Domains to Solve Function Optimization Problems

A new Ant Colony Optimization ACOB to solve function optimization problems (FOP) is evaluated experimentally by using ten standard multimodal test functions such as Michaelwicz's function. In ACOB, ants search for solutions in binary search space and can improve the accuracy of solutions by the stepwise localization of search space. Experiments show that ACOB can keep the balance between accuracy and efficiency to search for optimum solutions, and that it can reduce the population size of ACOR, which is a preceding ACO based on real search space. It is also shown that Covariance Matrix Adaptation-Evolution Strategy (CMA-ES) is superior in computational time but lacks the accuracy of solutions, and that Genetic Algorithm (GA) is superior in the ratio of getting the optimum solutions but weak in the performance.