Improving convergence properties of a differential evolution algorithm

The differential evolution is a popular and efficient way to solve complicated optimization tasks with many variables and constraints. In this article we study the ability of the differential evolution algorithms to attain the global minimum of the cost function. We demonstrate that although often d...

Full description

Saved in:
Bibliographic Details
Main Authors: Knobloch, Roman, Mlynek, Jaroslav, Srb, Radek
Format: Conference Proceeding
Language:English
Subjects:
Algorithms
Convergence
Design modifications
Evolutionary algorithms
Citations: Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The differential evolution is a popular and efficient way to solve complicated optimization tasks with many variables and constraints. In this article we study the ability of the differential evolution algorithms to attain the global minimum of the cost function. We demonstrate that although often declared as a global optimizer the classic differential evolution algorithm does not guarantee the convergence to the global minimum. To improve this weakness we design a modification of the classic differential evolution algorithm to enrich the diversity of its populations. This modification limits the premature convergence to local minima and ensures the asymptotic global convergence. We tested the modified algorithm in numerical experiments and compared the efficiency in finding the global minimum for the classic and modified algorithm. The modified algorithm is significantly more efficient with respect to the global convergence than the classic algorithm.
ISSN:0094-243X
1551-7616
DOI:10.1063/1.4968451