Application of design style in evolutionary multi-level networks synthesis

Abstract

Considers the evolutionary design of logical networks from the computer-aided design (CAD) point of view. Scanning the space of all possible network solutions by a scanning window is the the crucial point of an evolutionary paradigm. This is the basis for the implementation of CAD methods in order to improve recently obtained results on the evolutionary approach for network synthesis. Firstly, we introduce the concept of a target design style in evolutionary network synthesis and show that it is closely related to the CAD problem of multi-level network design over a fixed library of cells. Secondly, because the network search space is partitionable, we use the technique of switching function decomposition. Therefore, independent parallel processing of subspaces via genetic algorithms (GAs) is possible. Moreover, since GAs are inherently parallel, we achieve massively parallel processing. The experimental data demonstrate the efficiency of the proposed approach and large improvements over recently obtained results.