Open AccessDevelopment and validation of constraints handling in a Differential Evolution optimizer
Author(s) -
Mihai-Vlăduţ Hothazie,
Georgiana Cristina ICHIM,
Mihai Victor Pricop
Publication year - 2020
Publication title -
incas buletin
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.282
H-Index - 10
eISSN - 2247-4528
pISSN - 2066-8201
DOI - 10.13111/2066-8201.2020.12.1.6
Subject(s) - differential evolution , benchmark (surveying) , computer science , mathematical optimization , matlab , fidelity , differential (mechanical device) , scheme (mathematics) , convergence (economics) , optimization problem , genetic algorithm , code (set theory) , evolutionary computation , penalty method , algorithm , mathematics , machine learning , set (abstract data type) , engineering , telecommunications , mathematical analysis , geodesy , aerospace engineering , economic growth , economics , programming language , geography , operating system
Research work requires independent, portable optimization tools for many applications, most often for problems where derivability of objective functions is not satisfied. Differential evolution optimization represents an alternative to the more complex, encryption based genetic algorithms. Various packages are available as freeware, but they lack constraints handling, while constrained optimizations packages are commercially available. However, the literature devoted to constraints treatment is significant and the current work is devoted to the implementation of such an optimizer, to be applied in low-fidelity optimization processes. The parameter free penalty scheme is adopted for implementation, and the code is validated against the CEC2006 benchmark test problems and compared with the genetic algorithm in MATLAB. Our paper underlines the implementation of constrained differential evolution by varying two parameters, a predefined parameter for feasibility and the scaling factor, to ensure the convergence of the solution.