Fast minimization of Boolean functions with high minimality

Abstract

The authors describe a practical minimization procedure for Boolean functions with high minimality, PM1. The procedure is divided into two main modules: the clustering module which classifies the product terms in the primitive Boolean function into several disjoint groups containing fewer variables, and the minimizing module which maximizes the Boolean function corresponding to each of the groups by employing a machine learning algorithm, AE1. PM1 is a descendent version of SWT which has no clustering module. Experiments show that PM1 is more efficient than SWT, and more effective than ENTOTT which is only based on logic laws.<>