简介:Inthispaperweinvestigateseveralsolutionalgorithmsfortheconvexfea-sibilityproblem(CFP)andthebestapproximationproblem(BAP)respectively.Thealgorithmsanalyzedarealreadyknownbefore,butbyadequatelyreformulatingtheCFPortheBAPwenaturallydeducethegeneralprojectionmethodfortheCFPfromwell-knownsteepestdecentmethodforunconstrainedoptimizationandwealsogiveanaturalstrategyofupdatingweightparameters.Inthelinearcaseweshowtheconnec-tionofthetwoprojectionalgorithmsfortheCFPandtheBAPrespectively.Inaddition,weestablishtheconvergenceofamethodfortheBAPundermilderassumptionsinthelinearcase.WealsoshowbyexamplesaBauschke'sconjectureisonlypartiallycorrect.