简介:ANEFFECTIVEIMPROVEMENTOFREMESALGORITHMXiongguijing;(熊规景)Xiesiqing(谢四清)(Inst.ofMath.Sci.,ChineseAcademyofSciences,430071Wuhan,...
简介:Inthispaper,anewgloballyconvergentalgorithmfornonlinearoptimizationproblemswithequalityandinequalityconstraintsispresented.ThenewalgorithmisofSQPtypewhichdeterminesasearchdirectionbysolvingaquadraticprogrammingsubproblemperiteration.Somerevisionsonthequadraticprogrammingsubproblemhavebeenmadeinsuchawaythattheassociatedconstraintregionisnonemptyforeachpointxgeneratedbythealgorithm,i.e.,thesubproblemsalwayshaveoptimalsolutions.Thenewalgorithmhastwoimportantproperties.Thecomputationofrevisionparameterforguaranteeingtheconsistencyofquadraticsubproblemandthecomputationofthesecondordercorrectionstepforsuperlinearconvergenceusethesameinverseofamatrixperiteration,sothecomputationamountofthenewalgorithmwillnotbeincreasedmuchmorethanotherSQPtypealgorithms;Anotheristhatthenewalgorithmcangiveautomaticallyafeasiblepointasastartingpointforthequadraticsubproblemsperiteration,thiswillobivouslysimplifythecomputationprocedureofthesubproblems.Somenumericalresultsarereported.
简介:Inthispaperweconsidertheeffectivereducibilityofthefollowinglineardifferentialequation:x=(A+∈Q(t,∈))x,|∈|≤∈0,whereAisaconstantmatrix,Q(t,e)isquasiperiodicint,andeisasmallperturbationparameter.WeprovethatiftheeigenvaluesofAandthebasicfrequenciesofQsatisfysomenon-resonantconditions,thelineardifferentialequationcanbereducedtoy=(A^*(∈)+R^*(t,∈))y,|∈|≤∈o,whereR^*isexponentiallysmallin∈.