TY - GEN
T1 - On the fitness of high order schema of a linear-weighted coded genetic algorithm
AU - Mo, Hongqiang
AU - Li, Zhong
AU - Park, Jin Bae
AU - Joo, Young Hoon
AU - Li, Xiangyang
PY - 2008
Y1 - 2008
N2 - According to Schema Theorem, the larger the fitness value of a schema is, the higher the chance of the sub-space corresponding to the schema being chosen for searching is. Therefore, the coding of a genetic algorithm should be designed to produces short building blocks at as more fixed positions of the strings as possible. In this paper, an upper and lower bounds of the fitness values of order-n building blocks are given in terms of the maximal fitness values of order-1 building blocks. It is shown that, if a linear-weighted coding supplies order-1 building blocks at most of the loci of the strings, then there is sure to be remarkable fitness differences among high order schemata. In such case, the trajectory of the fitness function contains peaks, and diversity loss is likely to occur in the local search within these peaks.
AB - According to Schema Theorem, the larger the fitness value of a schema is, the higher the chance of the sub-space corresponding to the schema being chosen for searching is. Therefore, the coding of a genetic algorithm should be designed to produces short building blocks at as more fixed positions of the strings as possible. In this paper, an upper and lower bounds of the fitness values of order-n building blocks are given in terms of the maximal fitness values of order-1 building blocks. It is shown that, if a linear-weighted coding supplies order-1 building blocks at most of the loci of the strings, then there is sure to be remarkable fitness differences among high order schemata. In such case, the trajectory of the fitness function contains peaks, and diversity loss is likely to occur in the local search within these peaks.
UR - http://www.scopus.com/inward/record.url?scp=58149195463&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=58149195463&partnerID=8YFLogxK
U2 - 10.1142/9789812799470_0117
DO - 10.1142/9789812799470_0117
M3 - Conference contribution
AN - SCOPUS:58149195463
SN - 981279946X
SN - 9789812799463
T3 - World Scientific Proceedings Series on Computer Engineering and Information Science 1; Computational Intelligence in Decision and Control - Proceedings of the 8th International FLINS Conference
SP - 713
EP - 718
BT - World Scientific Proceedings Series on Computer Engineering and Information Science 1; Computational Intelligence in Decision and Control - Proceedings of the 8th International FLINS Conference
PB - World Scientific Publishing Co. Pte Ltd
T2 - Computational Intelligence in Decision and Control - 8th International FLINS Conference
Y2 - 21 September 2008 through 24 September 2008
ER -