Genetic Programmingwith Reuse of Known Designs for Industrially Scalable, Novel Circuit Design

Abstract

This paper shows how aggressive reuse of known designsbrings orders-of-magnitude reduction in computationaleffort, and simultaneously resolves trust issues forsynthesised designs, for genetic programming applied toautomated structural design. Furthermore, it usestrustworthiness tradeoffs to handle addition of noveltyin a trackable fashion. It uses a multi-objectivealgorithm with an age-layered population structure toavoid premature convergence. While the application hereis analog circuit design, the methodology is generalenough for many other problem domains.