Simplifying the Parameterization of Real-Coded Evolutionary Algorithms
Patrick M. Reed, Satoshi Yamaguchi · Critical Transitions in Water and Environmental Resources Management · 2004
This paper demonstrates how existing parameterization techniques for binary coded genetic algorithms can be extended to real-coded evolutionary algorithms. An easy-to-implement framework for automating parameter setting for real-coded evolutionary algorithms is demonstrated in this study using Differential Evolution (DE), a real-coded evolutionary algorithm. DE was selected because the algorithm has been successfully demonstrated on a wide range of applications as a result of its inclusion in both the Mathematica and MATLAB optimization toolboxes. Preliminary results show that the framework successfully eliminates trial-and-error parameter analysis, minimizes the algorithms' computational demands, and ensures that DE reliably attains high quality solutions. The parameterization framework is not application specific and can be easily implemented on other real-coded evolutionary algorithms.