Research on Continuous Ant Colony Optimization Algorithm and Application in Neural Network Modeling

Ant Colony Optimization (ACO) inspired by ant's foraging behavior is initially proposed for solving the combination optimization problems. In order to apply it to continuous domains, Krzysztof Socha and Marco Dorigo present an effective extension - ACOr - following the fundamental framework of ACO exactly. This paper explains the principle and mechanism of ACOr in detail. We test its performance using several typical benchmark functions and give the simulation results. Compared with earlier literatures about it, some additional analyses and comparisons with other heuristic evolutionary algorithms are shown. In addition, the continuous algorithm is applied successfully to neural network modeling for dynamical system. The results show that ACOr is a strong continuous optimization method and enriches the theory of ant algorithm.