lecture11

lecture11 - Yinyu Ye, MS&E, Stanford...

Info iconThis preview shows pages 1–5. Sign up to view the full content.

View Full Document Right Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Yinyu Ye, MS&E, Stanford MS&E211 Lecture Note #11 1 NonlinearOptimizationAlgorithmsII YinyuYe DepartmentofManagementScienceandEngineering StanfordUniversity Stanford,CA94305,U.S.A. http://www.stanford.edu/yyye Yinyu Ye, MS&E, Stanford MS&E211 Lecture Note #11 2 NonlinearOptimization:PrimalMethods ( NLP ) minimize f ( x ) subjectto g i ( x ) i =1 ,...,m. PrimalMethodsaremethodsthatworkontheoriginalproblemdirectlyby searchingthroughthe feasibleregionoftheproblem . Forsimplicity,assumethattheconstrainsarealllinear g i ( x ):= a i x- b i . Yinyu Ye, MS&E, Stanford MS&E211 Lecture Note #11 3 FeasibleDirectionMethods Searchalonga feasible directionthatisalso descent atthesametime. Lettheactiveconstraintsetbe A ( x k ):= { i : a i x k- b i =0 } ata feasible iterativepoint x k .Thenwesolve: ( FDM ) minimize f ( x k ) d subjectto a i d , i A ( x k ) k d k 1 . Supposetheminimizeris d k .Iftheminimalvalueis ,then x k isalreadya local minimizer ;otherwise x k +1 = x k + k d k where k isthe stepsize tomake x k +1 stay feasible and reduce theobjective functionthemost. Yinyu Ye, MS&E, Stanford MS&E211 Lecture Note #11 4 ActiveSetMethods Guesswhichconstraintswouldbe active attheminimizer,sayinaworkingset W k .Thenwesolve: ( ASM ) minimize f ( x ) subjectto a i x- b i =0 , i W k . TheminimizercanbecomputedfromtheKKTconditionsfor equality constrained problem(whichisgenerallyeasierthan inequality problemtosolve),andletitbe x k . If a i x k- b i < forsome i 6 W k ,then ADD thisviolatedconstraintinto W k andresolve(ASM);elseifoneoftheLagrangemultipliers i isnegative,then DROP theassociateconstraintfrom W k andresolve(ASM). Ifthereisnoneedto ADDorDROP , x k isa KKTpoint fortheoriginalinequality constrainedproblem. Yinyu Ye, MS&E, Stanford MS&E211 Lecture Note #11 5 OtherPrimalMethods Thegradientprojectionmethod :projectthegradientontothefeasibledirection space. Thereducedgradientmethod :likethesimplexmethodbychangingnon-basic variables. Theellipsoidmethod :Thebasicideasofthe ellipsoidmethod stemfromresearch doneinthenineteensixtiesandseventiesmainlyintheSovietUnion(asitwas thencalled).Theideainanutshellistoenclosetheregionofinterestineach memberofasequenceofellipsoidswhosesizeisdecreasing,resemblingthe bisection method. Yinyu Ye, MS&E, Stanford MS&E211 Lecture Note #11 6 LinearFeasibilityProblem Themethoddiscussedhereisreallyaimedatfindinganelementofa solutionset X givenbyasystemoflinearinequalities. X = { x R n : a i x- b i ,i =1 ,...m } Findinganelementof X canbethoughtofasbeing equivalent tosolvinga linear programming problem,thoughthisrequiresabitofdiscussion. Yinyu Ye, MS&E, Stanford MS&E211 Lecture Note #11 7 Twotechnicalassumptions (A1) X iscontainedbya ball centeredatthe origin witha radius R> . (A2) X hasa volume atleast n vol S ( , 1) . Here, vol S ( , 1) isthevolumeofthe unitball inthe n-dimensionalspace....
View Full Document

This note was uploaded on 05/06/2010 for the course MSE 211 at Stanford.

Page1 / 17

lecture11 - Yinyu Ye, MS&E, Stanford...

This preview shows document pages 1 - 5. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online