Approximation and Gradient Descent Training with Neural Networks

G. Welper

It is well understood that neural networks with carefully hand-picked weights provide powerful function approximation and that they can be successfully trained in over-parametrized…