Adaptive kernel principal components tracking.pdf

Adaptiveonlinealgorithmsforsimultaneouslyextractingnonlineareigenvectorsofkernelprincipalcomponentanalysis(KPCA)arede-veloped.KPCAneedsalltheobservedsamplestorepresentbasisfunctions,andthesamescaleofeigenvalueproblemasthenumberofsamplesshouldbesolved.ThispaperreformulatesKPCAanddeduces