Graph signal interpolation with positive definite graph basis functions

For the interpolation of graph signals with generalized shifts a basis function (GBF), we introduce concept positive definite functions on graphs. This merges kernel-based spectral theory graphs and can be regarded as analog radial in Euclidean spaces or spherical functions. We provide several descriptions graphs, most relevant one is Bochner-type characterization terms Fourier coefficients. These allow us to design GBF's study GBF more detail. are able characterize native interpolants. explicit estimates for error obtain bounds numerical stability. As final application, show how used get quadrature formulas