Evolving Graphs with Semantic Neutral Drift

TL;DR

This paper introduces Semantic Neutral Drift (SND) in genetic programming, leveraging equivalence laws to improve evolutionary performance through semantics-preserving mutations, especially in digital circuit design tasks.

Contribution

The paper presents a novel SND approach using equivalence laws for semantics-preserving mutations in GP, with empirical validation on digital circuit benchmarks.

Findings

SND leads to quantitative improvements in evolution.

Benefits of SND involve complex processes beyond simple growth.

Choosing detrimental parameters can be beneficial when using SND.

Abstract

We introduce the concept of Semantic Neutral Drift (SND) for genetic programming (GP), where we exploit equivalence laws to design semantics preserving mutations guaranteed to preserve individuals' fitness scores. A number of digital circuit benchmark problems have been implemented with rule-based graph programs and empirically evaluated, demonstrating quantitative improvements in evolutionary performance. Analysis reveals that the benefits of the designed SND reside in more complex processes than simple growth of individuals, and that there are circumstances where it is beneficial to choose otherwise detrimental parameters for a GP system if that facilitates the inclusion of SND.