Regularization Networks and Support Vector Machines
The connection between self-associated two-dimensional vector functionals and third degree forms
On the near optimality of the stochastic approximation of smooth functions by neural networks*