Network Structuring and Training Using Rule-based Knowledge

We demonstrate in this paper how certain forms of rule-based knowledge can be used to prestructure a neural network of normalized basis functions and give a probabilistic interpretation of the network architecture.