Nonparametric estimation of a probability density on a Riemannian manifold using Fourier expansions (Q923549): Difference between revisions

This paper bases on results of \textit{L. Devroye} and \textit{L. Györfi} [Nonparametric density estimation. The \(L_ 1\)-view. (1985; Zbl 0546.62015)] and ideas of \textit{E. Giné} [Ann. Stat. 3, 1243-1266 (1975; Zbl 0322.62058)]. The author assumed that the sample space is a closed or homogeneous m-dimensional Riemannian manifold and proposed a nonparametric density estimator of generalized orthonormal series type \(f^*\) for the s times differentiable density f. He proved \(L_ 2\) and \(L_{\infty}\) convergence rates: \[ E(\| f-f^*\|^ 2_{L_ 2})\leq O(N^{m(2s+m)^{-1}}N^{-1}), \] and \[ E(\| f-f^*\|^ 2_{L_{\infty}})\leq O(N^{2m(2s+m)^{-1}}N^{-1}). \]