傅 欣 欣
(重慶師范大學 數(shù)學學院,重慶 400030)
求解無約束全局優(yōu)化的一類單參數(shù)填充函數(shù)
傅 欣 欣
(重慶師范大學 數(shù)學學院,重慶 400030)
填充函數(shù)法是求解無約束全局極小化問題的一種有效方法,這種方法的關鍵是構造填充函數(shù).介紹了一種新的僅含有一個參數(shù)的填充函數(shù)和相應的填充函數(shù)算法,同時分析其填充性質;通過數(shù)值試驗證明了算法是有效的.
全局優(yōu)化;極小點;填充函數(shù)
考慮如下問題(P):
(1)
其中,f在Rn→R上連續(xù)可微.
假設1f(x)滿足強制條件:當‖x‖→+∞時,f(x)→+∞.
假設2 問題(P)的局部極小值的個數(shù)為有限個.由假設1可知:存在一個有界閉區(qū)間Ω?Rn,使得f(x)的全部全局極小點都包含在Ω的內部.因此,式(1)等價于求問題(P′):
minf(x)
s.t.x∈Ω
(P′)
下面給出文獻[2]中改進的填充函數(shù)定義:
對于問題(P′),提出一個新的單參數(shù)填充函數(shù):
證明如果x∈S1,則
若x∈S2,則
若{xk}?S1,則:
若{xk}?S2,則:
所以dTf(x),故時,dT所以,證畢.
證畢.
由以上的分析,下面給出填充函數(shù)算法.
主步:
(4) 若x2∈Ω,轉步驟(5);否則,令i:=i+1,轉步驟(3).
下面通過幾個算例來驗證其算法的有效性,算例均是在同一計算機用Matlab7.01編程進行運算,以下算例參見文獻[7].
(1) Two-dimensional函數(shù):全局最優(yōu)值f(x*)=4.402 3e-016.
(2) Goldstein and Price函數(shù):全局極小點x*=(0.000 0,-1.000 0)T,f(x*)=3.000 0.
(3) n-dimensional Sine-square函數(shù):n=5,全局極小點x*=(1.000 0,1.000 0,…,1.000 0)T,f(x*)=0.000 0.
(4) 6-hump back camel函數(shù):x*=(0.089 8,0.712 7)T,f(x*)=-1.041 2.
計算結果詳見表1.
表1 數(shù)值結果
[1] GE R P. A filled functions method for finding a global minimizer of a function of several variables[J]. Mathematical Programming,1990(46): 191-204
[2] GAO C L,YANG Y J,HUA B S. A new class of filled functions with one parameter for global optimization[J]. Computers and Mathematics with Applications,2011(62): 2393-2403
[3] YANG Y J,SHANG Y L. A new filled function method for unconstrained global optimization[J]. Applied Mathematics and Computation,2006(173):501-512
[4] ZHANG L S,NG C K,LI D. A new filled function method for global optimization[J]. Global Optim,2001(20):49-65
[5] LIU X. Finding global minimia with a computabal filled function[J]. Global Optim,2001(19):151-161
[6] LIU X. Several filled functions with mitigators[J]. Applied Mathematics and Computation,2002 (133):375-387
[7] LIANG Y M,ZHANG L S,LI M M,et al. A filled function method for global optimization[J]. Journal of Computational and Applied Mathematics,2007(205):16-31
Keywords:global optimization;minimum point;filled function
A Class of Single Parameter Filled Function for Solving Unconstrained Global Optimization
FUXin-xin
(School of Mathematics, Chongqing Normal University, Chongqing 400030, China)
The filled function method is an effective method to solve unconstrained global minimization problem and the key to this method is to construct filled function.This paper introduces a kind of new filled function with only one parameter and its corresponding filled function algorithm, meanwhile, analyzes its filling property and verifies that this algorithm is effective by numerical test.
1672-058X(2013)10-0021-04
2013-04-01;
2013-05-06.
傅欣欣(1988-),女,重慶人,碩士研究生,從事全局最優(yōu)化理論與算法研究.
O221.2
A
責任編輯:李翠薇