Partitioned incremental evolution of hardware using genetic programming

Abstract

In an effort to enable evolutionary computation techniques to discover solutions for large and complex hardware systems, techniques have been devised to break the initial problem down into smaller sub-tasks. In particular, a decomposition approach has been described that is based on partitioning of the circuit test vectors, but it has its limitations. In an effort to address this, we have combined the partitioning method with an incrementally evolving genetic programming approach. The result, referred to as Partitioned Incremental Evolution of HARDware (PIE-HARD), exhibits solution-finding performance that is significantly better than that of other approaches. © 2008 Springer-Verlag Berlin Heidelberg.