组织进化数值优化算法.pdfVIP

27 2 Vol . 27 No. 2 2004 2 CH INESE JOU RNA L O F COM PUT ERS F eb . 2004 刘 静 钟伟才 刘 芳 焦李成 ( 7 1007 1) , !! ! , . , , , , !! ! . . , 4 6 , 3 , . , . , , , . , . T P 18 An Organizational Evolutionary Algorithm for Constrained and Unconstrained Optimization Problems L IU J ing ZHON G WeiCai L IU F ang J IAO L iCheng ( K ey L abor atory f or R ad ar S ignal Processi ng , X id ian Univ ersi ty , X i∀an 7 1007 1) Abstract Based on the concept of organizat ion in econom ics, a novel evolut ionary algorit hm, Organi zational Evolut ionary A lgorit hm ( OEA ) , is proposed to deal w ith bot h unconst rained and const rained optimizat ion problems. It s mechanism is complet ely diff erent from t he traditional g enetic algorit hm, evolutionary programm ing and evolut ion strateg y . In OEA , evolutionary op erat ion s do not act on indi vidu als direct ly, but on organizat ion s, so t hree evolutionary operators is designed for organizat ions. T heoretical analysis proves th at OEA converges to t he glob al opt imum. In ex periments, OEA is tested on 4 uncon strained and 6 constrained benchm ark problems, and compared w ith three recent algo rithms. T he result s indicate th at OEA performs m uch bet ter t h an t he three ot her algorit hms bot h in the qualit y of solution and in the computational com plexit y . For const rained problems, good results are obtained by only using a very simple penalt y term, w hich show s th at OEA has a low probabilit y t o get trapped in the local opt ima and its sea