简介：Wehaveprototypedandanalyzeddesignofanovelapproachforthehighthroughputcomputing-acoreelementfortheemergingHENPcomputationalgrid.IndependenteventprocessinginHENPiswellsutedforcomputinginparallel.Theprototypefacilitatedsuseofinexpensivemass-marketcomponentsbypovidingfaulttolerantresilienece(insteadoftheexpensivetotalsystemreliablity)viahighlyscalablemanagementcomponents.TheabilitytohandlebothhardwareandsoftwarefailuresonalargededicatedHENPfacilitylimitstheneedforuserintervention.ArobustdatamanagementisespeciallyimportantinHENPcomputingsincelargedata-flowsoccurbeforeand/oratfereachprocessingtask.Thearchitectureofouractiveobjectobjectcoordinationschemaimplementsamulti-levelhierarchicalagentmodel,Itprovidesfaulttolerancebysplittingalargeoveralltaskintoindependentatomicprocesses,performedbylowerlevelagentssynchronizingeachotherviaalocaldatabase.Necessarycontrolfunctionperformedbyhigherlevelagentsinteractwiththesamedatabasethusmanagingdistributeddataproduction.ThesystemhasbeentestedinproductionenvironmentforsimulationsintheSTARexperimentatRHIC.Ourarchitecturalprototypecontrolledprocessesonmorethanahundredprocessorsatatimeandhasrunforextendedperiodsoftime.Twentyterabytesofsimulateddatahavabeenproduced.ThegenericnatureofourtwolevelarchitecturalsolutionfaulttoleranceindistributedenvironmenthasbeendemonstratedbyistsuccessfultestforthegridfilereplicationservicesbetweenBNLandLBNL.

简介：Thebilevelprogrammingisappliedtosolvehierarchicalintelligencecontrolproblemsinsuchfieldsasindustry,agriculture,transportation,military,andsoon.Thispaperpresentsaquadraticobjectivepenaltyfunctionwithtwopenaltyparametersforinequalityconstrainedbilevelprogramming.Undersomeconditions,theoptimalsolutiontothebilevelprogrammingdefinedbythequadraticobjectivepenaltyfunctionisprovedtobeanoptimalsolutiontotheoriginalbilevelprogramming.Moreover,basedonthequadraticobjectivepenaltyfunction,analgorithmisdevelopedtofindanoptimalsolutiontotheoriginalbilevelprogramming,anditsconvergenceprovedundersomeconditions.Furthermore,undertheassumptionofconvexityatlowerlevelproblems,aquadraticobjectivepenaltyfunctionwithoutlowerlevelproblemsisdefinedandisprovedequaltotheoriginalbilevelprogramming.

简介：Thispaperaimsatprovidinganuncertainbilevelknapsackproblem(UBKP)model,whichisatypeofBKPsinvolvinguncertainvariables.AndthenanuncertainsolutionfortheUBKPisproposedbydefiningPENashequilibriumandPEStackelbergNashequilibrium.Inordertoimprovethecomputationalefficiencyoftheuncertainsolution,severaloperators(binarycodingdistance,inversionoperator,explosionoperatorandbinarybacklearningoperator)areappliedtothebasicfireworksalgorithmtodesignthebinarybackwardfireworksalgorithm(BBFWA),whichhasagoodperformanceinsolvingtheBKP.Asanillustration,acasestudyoftheUBKPmodelandthePEuncertainsolutionisappliedtoanarmamentstransportationproblem.

简介：Withthedeepenofmarketcompetition,productpricingandproductiondecisionprobleminmanyfirmshavebecomemoreandmoreimportant.Abilevelmodelisproposedtodescribethepricingandproductiondecisionswithfuzzydemandandfuzzycostparameters.Theupperlevelistodeterminetheoptimalpriceandproductionquantitywithcapacityconstraints.Usingthisinformation,thelowerlevelproblemtriestostructurearesponse(thedistributionpatternofcustomers(ormarkets))thatwillsatisfyhisdemandatminimumcost.AndaftertransformingthefuzzynumbersintothecrispvaluebyGradedMeanIntegrationRepresentationmethod,thesolutionalgorithmbasedondifferencemethodisgiven.Finally,theapplicationofthemodelanditsalgorithmareillustratedwithasimpleexample.

简介：Adiscretedifferentialevolutionalgorithmcombinedwiththebranchandboundmethodisdevelopedtosolvetheintegerlinearbilevelprogrammingproblems,inwhichbothupperlevelandlowerlevelvariablesareforcedtobeinteger.Anintegercodingforupperlevelvariablesisadopted,andthenadiscretedifferentialevolutionalgorithmwithanimprovedfeasibility-basedcomparisonisdevelopedtodirectlyexploretheintegersolutionattheupperlevel.Foragivenupperlevelintegervariable,thelowerlevelintegerprogrammingproblemissolvedbytheexistingbranchandboundalgorithmtoobtaintheoptimalintegersolutionatthelowerlevel.Inthesameframeworkofthealgorithm,twootherconstrainthandlingmethods,i.e.thepenaltyfunctionmethodandthefeasibility-basedcomparisonmethodarealsotested.Theexperimentalresultsdemonstratethatthediscretedifferentialevolutionalgorithmwithdifferentconstrainthandlingmethodsiseffectiveinfindingtheglobaloptimalintegersolutions,buttheimprovedconstrainthandlingmethodperformsbetterthantwocomparedconstrainthandlingmethods.

简介：Anewbilevelgeneralizedmixedequilibriumproblem(BGMEP)isintroducedandstudiedintopologicalvectorspaces.Byusingaminimaxinequality,theexistenceofsolutionsandthebehaviorofsolutionsetfortheBGMEParestudiedunderquitemildconditions.Theseresultsarenewandgeneralizesomerecentresultsinthisfield.

简介：Thispaperproposesasufficientcondition,ifthepropertiesofthebilevelprogrammingsatisfiedthecondition,wecansolvethebilevelprogrammingbysolvingsinglelevelprogramming.Itbecomeseasyandsimpletosolvethebilevelprogramming.

简介：Forill-posedbilevelprogrammingproblem,theoptimisticsolutionisalwaysthebestdecisionfortheupperlevelbutitisnotalwaysthebestchoiceforbothlevelsiftheauthorsconsiderthemodel'ssatisfactorydegreeinapplication.Toacquireamoresatisfyingsolutionthantheoptimisticonetorealizethetwolevels'mostprofits,thispaperconsidersbothlevels'satisfactorydegreeandconstructsaminimizationproblemofthetwoobjectivefunctionsbyweightedsummation.Then,usingthedualitygapofthelowerlevelasthepenaltyfunction,theauthorstransferthesetwolevelsproblemtoasingleoneandproposeacorrespondingalgorithm.Finally,theauthorsgiveanexampletoshowamoresatisfyingsolutionthantheoptimisticsolutioncanbeachievedbythisalgorithm.

简介：Thispaperproposesanonmonotonicbacktrackingtrustregionalgorithmviabilevellinearprogrammingforsolvingthegeneralmulticommodityminimalcostflowproblems.Usingthedualitytheoryofthelinearprogrammingandconvextheory,thegeneralizeddirectionalderivativeofthegeneralmulticommodityminimalcostflowproblemsisderived.Theglobalconvergenceandsuperlinearconvergencerateoftheproposedalgorithmareestablishedundersomemildconditions.

简介：Anewbilevelgeneralizedmixedequilibriumproblem(BGMEP)involvinggeneralizedmixedvariational-likeinequalityproblems(GMVLIPs)isintroducedandstudiedinthereflexiveBanachspaces.First,anauxiliarygeneralizedmixedequilibriumproblem(AGMEP)isintroducedtocomputetheapproximatesolutionsoftheBGMEPinvolvingtheGMVLIPs.Byusingaminimaxinequality,theexistenceandtheuniquenessofsolutionsoftheAGMEPareprovedundermildconditionswithoutanycoerciveassumptions.Byusinganauxiliaryprincipletechnique,thenewiterativealgorithmsareproposedandanalyzed,withwhichtheapproximatesolutionsoftheBGMEParecomputed.Thestrongconvergenceoftheiterativesequencegeneratedbythealgorithmsisshownundermildconditionswithoutanycoerciveassumptions.Thesenewresultscangeneralizesomerecentresultsinthisfield.