The aim of this work is towards a better understanding of the effect of using constraint violations in guiding evolutionary search for nonlinear programming problems. Different penalty functions, based on constraint violations, create different search biases. However, this bias may be eliminated when treating the nonlinear programming problem as a multiobjective task. The different search behaviors are illustrated using a new artificial test function. The effectiveness of the multiobjective approach is also compared with the standard penalty function method on a number of commonly used benchmark problems. It is shown that in practice multiobjective methods are not an efficient or effective approach to constrained evolutionary optimization.
[1]
H. T. Kung,et al.
On Finding the Maxima of a Set of Vectors
,
1975,
JACM.
[2]
Xin Yao,et al.
Stochastic ranking for constrained evolutionary optimization
,
2000,
IEEE Trans. Evol. Comput..
[3]
Zbigniew Michalewicz,et al.
Test-case generator for nonlinear continuous parameter optimization techniques
,
2000,
IEEE Trans. Evol. Comput..
[4]
Carlos A. Coello Coello,et al.
THEORETICAL AND NUMERICAL CONSTRAINT-HANDLING TECHNIQUES USED WITH EVOLUTIONARY ALGORITHMS: A SURVEY OF THE STATE OF THE ART
,
2002
.
[5]
Thomas Philip Runarsson,et al.
Reducing Random Fluctuations in Mutative Self-adaptation
,
2002,
PPSN.