New structures such the radial basis function network

38

Perceptron is a single layer NN, hence it has no hidden layer.

39

In the late 80s to early 90s, we have seen many developments in NNs. New NN structures such as: the radial basis function network [41], bidirectional associative memories [126], B-splines network [164], digital neurocomputer [97], functional-link networks [181], knapsack packing networks [98] and wavelet networks [248] were proposed. Different combinations of network topologies to improve the rep-resentational structure, such as recurrent backpropagation which combines feedback and feedforward topologies, were explored [3, 188, 240].

Figure 3.2: The basic elements of neural networks.

Summer