Bi-Decompositions of Multi-Valued Functions for Circuit Design and Data Mining Applications

We present eÆcient algorithms for the bi-decomposition of arbitrary incompletely speci ed functions in variable-valued logic. Several special cases are discussed. The algorithms are especially applicable for Data Mining applications, because, in contrast to the general multi-valued approaches to function decomposition that decompose to arbitrary tables, we create a network from multi-valued two-input operators that are selected by the user. Such decompositions lead to decision rules that are easier to understand by humans.