简介:Thepaperpresentsaneuralnetworkforsolvingaclassofquadraticprogrammingproblems.Theneuralnetworkiscompletelystabletoexactsolutionsandtherearenoparameterstoset.Moreover,noanaloguemultipliersanddividersarerequired,incontrasttoexistingneuralnetwork[3]whichneedsplentyofanaloguemultipliers.
简介:Thebasicprobleminoptimizingcommunicationnetworksistoassignapropercircuitforeachorigindestinationpairinnetworkssoastominimizetheaveragenetworkdelay,andthenetworkoptimalrouteselectionmodelisamulti-constrained0-1nonlinearprogrammingproblem,Inthispaper,anewstochasticoptimizationalgorthm,ImmuneAlgorithm,isappliedtosolvetheoptimizationproblemincommunicationnetworks,AndthebackbonenetworkvBNSischosentoillustratethetechniqueofevaluatingdelayinavirtualnetowrk.Atlast,IAiscomparedwiththeoptimizationmethodincommunicationnetworksbasedonGeneticAlgorithm,andtheresultshowsthatIAisbetterthanGAinglobaloptimumfinding.