Abstract
The study introduces a new class of fuzzy neurons and fuzzy neural networks exploiting a model of a generalized multivalued exclusive-OR (XOR) operation. The proposed neural architecture is useful in an algebraic representation (description) of fuzzy functions regarded as mappings between unit hypercubes, say [0, 1](n), [0, 1](m). Some underlying properties of the fXOR neurons are discussed and a detailed learning algorithm is given along with a number of illustrative numeric examples.