资源描述：
《A Robust Algorithm for Solving Nonlinear Programming Problems》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。
1、ARobustAlgorithmforSolvingNonlinearProgrammingProblemsYanLiLi-ShanKangComputationCenter,StateKeyLaboratoryofSoftwareEngineering,WuhanUniversity,Wuhan,430072,ChinaWuhanUniversity,Wuhan,430072,Chinallyyan2000@21cn.comAbstractweintroducetheconceptofasubspaceVofthedomainD.Inthispap
2、er,weintroduceanewalgorithmforsolvingnonlinearmpoints(Xj,Yj),j=1,2,L,minDareusedtoprogramming(NLP)problems.ItisanextensionofGuo’salgorithm[1]constructthesubspaceV,definedas-mwhichpossessesenhancedcapabilitiesforsolvingNLPproblems.TheseV={(Xv,Yv)ÎD(Xv,Yv)=åi=1ai(Xi,Yi)}capabilit
3、iesinclude:a)advancingthevariablesubspace,b)addingamwhereaiissubjecttoåi=1ai=1,-0.5£ai£1.5.searchprocessoversubspacesandnormalizedconstraints,c)usinganBecauseGuo’salgorithmdealsmainlywithoptimizationadaptivepenaltyfunction,andd)addingtheabilitytodealwithintegerproblemswhichhave
4、realvariablesandINequalityNLPproblems,0-1NLPproblems,andmixed-integerNLPproblemsconstraints,weassumek1=0andq=0intheexpression(1).whichhaveequalityconstraints.Thesefourenhancementsincreasetheì0,gi(X)£0Denotingwi(X)=ícapabilitiesofthealgorithmtosolvenonlinearprogrammingproblemsîg
5、i(X),otherwiseinamorerobustanduniversalway.Thispaperwillpresentresultsofk2numericalexperimentswhichshowthatthenewalgorithmisnotonlyandW(X)=åwi(X)i=1morerobustanduniversalthanitscompetitors,butalsoitsWedefineaBooleanfunction“better”as:performancelevelishigherthananyothersintheli
6、terature.ìW(X1)£W(X2)TRUEïïW(X1)>W(X2)FALSEbetter(X,X)=í12ï(W(X1)=W(X2))Ù(f(X1)£f(X2))TRUEï(W(X1)=W(X2))Ù(f(X1)>f(X))FALSEî2Ifbetter(X1,X2)isTRUE,thismeansthattheindividualX1is1INTRODUCTIONTOGUO’SALGORITHM“better”thantheindividualX2.Guo’salgorithmcanbedescribedasfollows:Itwassh
7、ownin[1]thatGuo’salgorithmhasmanyGuo’sAlgorithmadvantageswhensolvingoptimizationproblems.SincetheBeginnewalgorithmpresentedinthispaperisbaseduponGuo’sinitializepoplnP={X1,X2,…,XN};XiÎDalgorithm,itisadvisabletofirstpresentbrieflyitsmain/since(q=0impliesnointegervariables)/genera
8、tioncountt:=0;features.Xbest=argMinf(Xi);ThegeneralNLP
