ABSTRACT
Differential Evolution Strategy (DES) is a method that combines the differential mutation with the search direction adaptation mechanisms used by the Covariance Matrix Adaptation Evolution Strategy (CMA-ES). Although earlier research on that algorithm proved its good efficiency, it was still outperformed by the combined and hybrid methods which have been the winners of single objective bound constrained numerical optimization competitions. This paper reports on research that was aimed at improving the efficiency of DES in such a way that the optimization process is initially performed by DES, and after it terminates, the result is finely tuned by CMA-ES, whose expectation vector and the covariance matrix are initialized with statistics of points generated by DES. The hybrid method is evaluated according to the problem definitions and evaluation criteria for the 2022 Special Session and Competition on Single Objective Bound Constrained Numerical Optimization. According to the numerical results, the proposed hybrid method outperforms the standard versions of both DES and CMA-ES. Moreover, the comparison of results on the CEC'2017 benchmark suite evidences that the presented method would be superior or comparable to other methods whose results for CEC'2017 have been reported by the competing teams.
- Ali Wagdy Mohamed Anas A. Hadi P.N. Suganthan Abhishek Kumar, Kenneth V. Price. 2020. Problem Definitions and Evaluation Criteria for the CEC 2022 Special Session and Competition on Single Objective Bound Constrained Numerical Optimization. Technical Report. Nanyang Technol. Univ., Singapore.Google Scholar
- Jarosław Arabas and Dariusz Jagodziński. 2020. Toward a Matrix-Free Covariance Matrix Adaptation Evolution Strategy. IEEE Transactions on Evolutionary Computation 24, 1 (2020), 84--98.Google ScholarDigital Library
- N. H. Awad, M.Z. Ali, J.J. Liang, B.Y. Qu, and Ponnuthurai N Suganthan. 2016. Problem definitions and evaluation criteria for the CEC 2017 special session and competition on real-parameter optimization. Technical Report. Nanyang Technol. Univ., Singapore and Jordan Univ. Sci. Technol. and Zhengzhou Univ., China.Google Scholar
- R. Biedrzycki. 2017. A version of IPOP-CMA-ES algorithm with midpoint for CEC 2017 single objective bound constrained problems. In IEEE Congr. Evol. Comput. 1489--1494.Google ScholarDigital Library
- Jakob Bossek. 2016. cmaesr: Pure R implementation of the Covariance Matrix Adaptation - Evolution Strategy (CMA-ES) with optional restarts (IPOP-CMA-ES). https://cran.r-project.org/web/packages/cmaesr/ R package version 1.0.3.Google Scholar
- Nikolaus Hansen. 2006. The CMA evolution strategy: a comparing review. In Towards a new evolutionary computation: advances on estimation of distribution algorithms, Jose A Lozano (Ed.). Springer, 75--102.Google Scholar
- Nikolaus Hansen. 2017. A Practical Guide to Benchmarking and Experimentation. In Proceedings of the Genetic and Evolutionary Computation Conference Companion (Berlin, Germany) (GECCO '17). Association for Computing Machinery, New York, NY, USA, 413. Google ScholarDigital Library
- Dariusz Jagodziński and Jaroslaw Arabas. 2017. A differential evolution strategy. In IEEE Congr. Evol. Comput. 1872--1876.Google ScholarDigital Library
- P.N. Suganthan. [n.d.]. https://github.com/P-N-Suganthan.Google Scholar
- Eryk Warchulski. 2022. cecs: R Interface for the C Implementation of CEC Benchmark Functions. https://github.com/ewarchul/cecs R package version 0.2.4.Google Scholar
Index Terms
- Improving the differential evolution strategy by coupling it with CMA-ES
Recommendations
DE/BBO: a hybrid differential evolution with biogeography-based optimization for global numerical optimization
Differential evolution (DE) is a fast and robust evolutionary algorithm for global optimization. It has been widely used in many areas. Biogeography-based optimization (BBO) is a new biogeography inspired algorithm. It mainly uses the biogeography-based ...
Self-adaptive differential evolution algorithm with improved mutation strategy
Different mutation strategies and control parameters settings directly affect the performance of differential evolution (DE) algorithm. In this paper, a self-adaptive differential evolution algorithm with improved mutation strategy (IMSaDE) is proposed ...
A novel similarity-based mutant vector generation strategy for differential evolution
GECCO '18: Proceedings of the Genetic and Evolutionary Computation ConferenceThe mutant vector generation strategy is an essential component of Differential Evolution (de), introduced to promote diversity, resulting in exploration of novel areas of the search space. However, it is also responsible for promoting intensification, ...
Comments