Fuzzy neural networks with reference neurons as pattern classifiers

A heterogeneous neural network consisting of logic neurons and realizing mappings in (0, 1) hypercubes is presented. The two kinds of neurons studied are utilized to perform matching functions (equality or reference neurons) and aggregation operations (aggregation neurons). All computations are driven by logic operations widely used in fuzzy set theory. The network is heterogeneous in its nature and includes two types of neurons organized into a structure detecting individual regions of patterns (using reference neurons) and combining them to yield a final classification decision.<
>