Extensions of a Theory of Networks for Approximation and Learning: Outliers and Negative Examples