摘要
Thispaperpresentsnonmonotonicquasi-Newtonalgorithmsviatwopre-conditionalcurvilinearpaths,thepreconditionalmodifiedgradientpathandtheprecon-ditionaloptimalpath,forunconstrainedoptimizationproblem.WeemploythestableBunch-Parlettfactorizationmethodtoformtwocurvilinearpathsveryeasily.Thenonmonotonecriterionisusedtospeeduptheconvergenceprogressinthecontoursofobjectivefunctionwithlargecurvature.Theoreticalanalysesaregivenwhichprovethattheproposedalgorithmsaregloballyconvergentandhavealocalsuperlinearcon-vergencerateundersomereasonableconditions.Theresultsofnumericalexperimentsarereportedtoshowtheeffectivenessoftheproposedalgorithms.
出版日期
2003年01月11日(中国期刊网平台首次上网日期,不代表论文的发表时间)