【正文】
不可微非線性函數(shù)優(yōu)化問題具有廣泛的工程和應用背景,如結(jié)構(gòu)設 計中使 結(jié)構(gòu)內(nèi)最大應力最小而歸結(jié)為極大極小優(yōu)化( minmax)問題、數(shù)據(jù)魯棒性擬合中采取最小絕對值準則建立失擬函數(shù)等。數(shù)值算例對混合方法的有效性進行了驗證。 fi’ = af i’ b ( fi 179。 step6 對每個個體 按照概率 pPowell進行 Powell搜索。 (3) 求 使得 =min ,令 = = ,若 ,則 Powell方法計算結(jié)束,得點 ;否則,執(zhí)行 (4)。二進制編碼比浮點編碼遺傳算法計算精度低,對于標準遺傳算法以目標函數(shù)小于 800為搜索成功,標準遺傳算法運行 100次。計算結(jié)果表明混合法優(yōu)于遺傳算法和 Powell法,可以可靠地搜索到具有多個局部極值的函數(shù)優(yōu)化問題的全局解。 Hybrid approach for global optima of indifferentiable nonlinear function Abstract A hybrid putational intellective algorithm for locating the global optima of indifferentiable nonlinear function was put forward by setting the Powell algorithm in realcode geic algorithm. The hybrid approach improved the local searching ability of the geic algorithm and promoted the probability for the global optima greatly. Because only the objective values are used, the hybrid approach is a generalized geic algorithm for global optima of differentiable and indifferentiable nonlinear functions. Key words global optima; hybrid approach; geic algorithms; Powell algorithm