Thispaperproposesasupervisedtraining-testmethodwithGeneticProgramming(GP)forpatternclassification.Comparedandcontrastedwithtraditionalmethodswithregardtodeterministicpatternclassifiers,thismethodistrueforbothlinearseparableproblemsandlinearnon-separableproblems.Forspecifictrainingsamples,itcanformulatetheexpressionofdiscriminatefunctionwellwithoutanypriorknowledge.Atlast,anexperimentisconducted,andtheresultrevealsthatthissystemiseffectiveandpractical.
