Interpretation of nodes in networks for fuzzy logic

The authors introduce a method to analyze the operation of individual nodes in a neural network where the nodes implement weighted Yager additive hybrid operators. This is because, after training, the neurons can be viewed as mini-rules which are (primarily) conjunctions, disjunctions, or compensators, and where the resulting weights have been shown to indicate the relative importance of the piece of evidence. It is shown that these nodes can be trained to give satisfying results for simple cases of fuzzy logic inference