Theessenceofthelinearsearchisone-dimensionnonlinearminimizationproblem,whichisanimportantpartofthemulti-nonlinearoptimization,itwillbespendthemostofoperationcountforsolvingoptimizationproblem.Toimprovetheefficiency,wesetaboutfromquadraticinterpolation,combinetheadvantageofthequadraticconvergencerateofNewton'smethodandadopttheideaofAnderson-Bjorckextrapolation,thenwepresentarapidlyconvergencealgorithmandgiveitscorrespondingconvergenceconclusions.Finallywedidthenumericalexperimentswiththesomewell-knowntestfunctionsforoptimizationandtheapplicationtestoftheANNlearningexamples.Theexperimentresultsshowedthevalidityofthealgorithm.
