Hybrid Neural Nets
James J. Buckley, Thomas Feuring · Studies in fuzziness and soft computing · 1998
The objective of this chapter is to show how to construct hybrid neural nets ( HNN ) to be computationally identical to discrete fuzzy expert systems (discussed in Chapter 4) and certain fuzzy controllers. An example of a 2 — 3 — 1 HNN is shown in Figure 5.1. Figure 5.1 is similar to Figure 3.1, however, the output y will be computed differently. As before (Chapter 3) the input neurons simply distribute the input signals to all the neurons in the middle layer. The HNN combines the signals and the weights using a t -norm T , and then aggregate the results, over all incoming arcs to a neuron, using a t -conorm C .