ABSTRACT
Inhibitory connections in neural networks serve a variety of purposes. In our discussion of random nets (Section 2.2), we noted that inhibition can facilitate the stabilization of network activity levels. Also in our discussion of network principles (Chapter 1), we noted that inhibition can provide a mechanism for making choices. These choices might be, for example, between input patterns for short-term memory storage, between categories for classification of a single input pattern, or between drives for activation. It must be added, however, that the choices are not always all-or-none.
