Eignets for function approximation on manifolds (Q984660): Difference between revisions

From the abstract: Let \(\mathbb{X}\) be a compact, smooth connected, Riemannian manifold without boundary, \(G:\mathbb{X}\to \mathbb{X}\to\mathbb{R}\) be a kernel. Analogous to a radial basis function network, an eignet is an expression of the form \(\sum^M_{j=1} \alpha_jG(\circ,y_j)\), where \(a_j\in \mathbb{R}\), \(y_j\in\mathbb{X}\), \(1\leq j\leq M\). A deterministic, universal algorithm for constructing an eignet for approximating functions in \(L^p(\mu;\mathbb{X})\) for a general class of measures \(\mu\) and kernels \(G\) is described. The algorithm yields linear operators. Using the minimal separation among the centers \(y_j\) as the cost of approximation, the modulus of smoothness estimates for the degree of approximation by our eignets is given and it's shown by means of a converse theorem that these are the best possible for every individual function. The estimates on the coefficients \(a_j\) in terms of the norm of the eignet are given also. Finally, it's demonstrated that if any sequence of eignets satisfies the optimal estimates for the degree of approximation of a smooth function, measured in terms of the minimal separation, then the derivatives of the eignets also approximate the corresponding derivatives of the target function in an optimal manner.