Combinational Circuit Design with Estimation of Distribution Algorithms

Abstract

The authors introduce new approaches for the combinational circuit design based on Estimation of Distribution Algorithms. In this paradigm, the structure and data dependencies embedded in the data (population of candidate circuits) are modeled by a conditional probability distribution function. The new population is simulated from the probability model thus inheriting the dependencies. The authors explain the procedure to build an approximation of the probability distribution through two approaches: polytrees and Bayesian networks. A set of circuit design experiments is performed and a comparison with evolutionary approaches is reported.