Hermite interpolation by neural networks

In this paper we study Hermite interpolation by single-layer feedforward networks. We provide upper bound estimates for the minimum number of hidden units that can solve the Hermite interpolation problem exactly. We consider both the univariate and the multivariate cases.