Title | ||
---|---|---|
Using differential evolution strategies in chemical reaction optimization for global numerical optimization. |
Abstract | ||
---|---|---|
In this paper we propose a new hybrid metaheuristic approach which combines Chemical Reaction Optimization and Differential Evolution to solve global numerical optimization problems. Chemical Reaction Optimization is widely used in several optimization problems. However, due to its random behavior in searching the optimal solution, it may converge slowly. Differential Evolution is another efficient method based on differentiation operation which can be achieved by several, more or less selective, research strategies. The aim of this paper is to propose new hybrid algorithms that use Differential Evolution strategies inside Chemical Reaction Optimization process in order to overcome its limits by increasing optimal quality and accelerating convergence. We propose in this paper two new hybrid algorithms. Both of them use the Differential Evolution Best Strategy as a local search operator to improve the exploitation process and the Differential Evolution Random Strategy as a global search operator to maintain the diversity of the population and improve the exploration process. However, the two proposed algorithms slightly differ on the used local search operators. Based on 23 benchmark functions classified in 3 categories, experimental studies start by showing that our second proposed algorithm is better than the first one. Then, this second algorithm is compared with numerous other existing algorithms. First, the experimental results of comparison with the original algorithms show that our algorithm attains very good performance for (1) the quality of the obtained solutions, where it outperforms the other algorithms by achieving the first average and overall rank for two over the three categories; (2) for the robustness where it obtains the best average number of successful runs (21.47 over 25 runs) as well as for (3) convergence speed where our proposed algorithm converges faster comparing with other algorithms in nine over the twenty three functions and finds better solution for functions where other algorithms converge faster. In addition, the proposed algorithm has also been compared with other hybrid chemical reaction and differential evolution based algorithms, the experimental results show that globally the proposed algorithm also outperforms the other hybrid algorithms except for some limited cases. |
Year | DOI | Venue |
---|---|---|
2017 | https://doi.org/10.1007/s10489-017-0921-4 | Appl. Intell. |
Keywords | Field | DocType |
Global numerical optimization,Metaheuristics,Hybrid algorithm,Chemical reaction optimization,Differential evolution | Hybrid algorithm,Computer science,Artificial intelligence,Optimization problem,Metaheuristic,Mathematical optimization,Test functions for optimization,Meta-optimization,Algorithm,Differential evolution,Multi-swarm optimization,Local search (optimization),Machine learning | Journal |
Volume | Issue | ISSN |
47 | 3 | 0924-669X |
Citations | PageRank | References |
3 | 0.37 | 28 |
Authors | ||
2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Mourad Nouioua | 1 | 5 | 2.10 |
Zhiyong Li | 2 | 64 | 11.15 |