Feedforward networks with one hidden layer and their rates of approximation
Kateřina Hlaváčková · 1998
We present an overview of some rates of approximation with respect to various computational units in the hidden layer of a one-layered feedforward neural network. The problem of estimating the number of units in the hidden layer is examined to ensure a given degree of approximation for a given function class. The results are discussed in terms of their complexity and functional characteristics.