Directed variation in evolution strategies
Qing Zhou
Yanda Li
Dept. of Autom., Tsinghua Univ., Beijing, China;
This paper appears in: Evolutionary Computation, IEEE Transactions on
Publication Date: Aug. 2003
Volume: 7,
Issue: 4
On page(s): 356- 366
ISSN: 1089-778X
INSPEC Accession Number: 7733142
Digital Object Identifier: 10.1109/TEVC.2003.812215
Current Version Published: 2003-08-26
Abstract
Biological evolution gives rise to self-organizing phenomena. Inspired by this theory, directed variation is added to the (μ, λ) evolution strategies (ES) algorithm and it is called directed variation ES (DVES). In DVES, some neighboring individuals in the population mutate correlatively according to the distribution of the whole population. Experimental results showed that, with the same number of function evaluations, directed variation ES reached better optimization results for different generally used strategies under the ES framework. Experimental analysis showed that the application of directed variation could increase the expected fitness improvement and the probability of fitness improvement. From a biological perspective, directed variation can be regarded as a result of self-organizing evolution.
Index
Terms
Available to subscribers and IEEE members.
References
Available to subscribers and IEEE members.
Citing Documents
Available to subscribers and IEEE members.
Cart
