Improving linearity of position-sensitive detector using support vector machines

Anintelligentmethodforimprovingpositionlinearityofposition-sensitivedetector(PSD),basedonsupportvectormachines(SVMs),isdeveloped.TheSVMisestablishedbasedonthestructuralriskminimizationprincipleratherthanminimizingtheempiricalerrorcommonlyimplementedinneuralnetworks.SVMcanachievehighergeneralizationperformance.TrainingSVMisequivalenttosolvingalinearlyconstrainedquadraticprogrammingproblem,thusthesolutionofSVMisalwaysuniqueandgloballyoptimal.Theimprovingpositionlinearityprocedurehasbeenillustratedusingatwo-dimensional(2D)PSD.ItispointedoutthatthepositionlinearityofthemeasuringsystemwithaproperSVMcorrectionisimprovedbytwoordersofmagnitudeinthemeasurementrange.