LE QUY DON
Technical University
VietnameseClear Cookie - decide language by browser settings

DEAL: A direction-guided evolutionary algorithm

Vu, C.C. and Bui, L.T. and Abbass, H.A. (2012) DEAL: A direction-guided evolutionary algorithm. In: 9th International Conference on Simulated Evolution and Learning, SEAL 2012, 16 December 2012 through 19 December 2012, Hanoi.

Text
A direction-guided evolutionary algorithm.pdf

Download (354kB) | Preview

Abstract

In this paper, we propose a real-valued evolutionary algorithm being guided by directional information. We derive direction of improvement from a set of elite solutions, which is always maintained overtime. A population of solutions is evolved over time under the guidance of those directions. At each iteration, there are two types of directions that are being generated: (1) convergence direction between an elite solution (stored in an external set) and a second-ranked solution from the current population, and (2) spreading direction between two elite solutions in the external set. These directions are then used to perturb the current population to get an offspring population. The combination of the offsprings and the elite solutions is used to generate a new set of elite solutions as well as a new population. A case study has been carried out on a set of difficult problems investigating the performance and behaviour of our newly proposed algorithm. We also validated its performance with 12 other well-known algorithms in the field. The proposed algorithm showed a good performance in comparison with these algorithms. © 2012 Springer-Verlag.

Item Type: Conference or Workshop Item (Paper)
Divisions: Faculties > Faculty of Information Technology
Identification Number: 10.1007/978-3-642-34859-4_15
Uncontrolled Keywords: Direction of improvement; Directional information; Evolutionary algorithms; Iterative methods
Additional Information: Conference code: 94550. Language of original document: English.
URI: http://eprints.lqdtu.edu.vn/id/eprint/10084

Actions (login required)

View Item
View Item