Modelling biological evolvability: implicit context and variation filtering in enzyme genetic programming

Abstract

This paper describes recent insights into the role of implicit context within the representations of evolving artefacts and specifically within the program representation used by enzyme genetic programming. Implicit context occurs within self-organising systems where a component’s connectivity is both determined implicitly by its own definition and is specified in terms of the behavioural context of other components. This paper argues that implicit context is an important source of evolvability and presents experimental evidence that supports this assertion. In particular, it introduces the notion of variation filtering, suggesting that the use of implicit context within representations leads to meaningful variation filtering whereby inappropriate change is ignored and meaningful change is encouraged during evolution.

Introduction

Enzyme genetic programming Lones and Tyrrell, 2001a, Lones and Tyrrell, 2001b, Lones and Tyrrell, 2002b, Lones and Tyrrell, 2004 is a form of genetic programming (GP) (Koza, 1992) which uses a program representation modelled upon biological enzyme systems. The approach is motivated both by the limitations of conventional GP representations and by the premise that biological representations are well adapted for representing entities undergoing evolutionary processes, irrespective of whether these entities are biological or non-biological in nature. The logic behind this reasoning has been addressed in earlier papers on enzyme GP (e.g. Lones and Tyrrell, 2001b). These papers also present comparative analysis of the performance of enzyme GP (Lones and Tyrrell, 2002a), the development of the enzyme model Lones and Tyrrell, 2001b, Lones and Tyrrell, 2002b and the evolution of solution size Lones and Tyrrell, 2002a, Lones and Tyrrell, 2002b. The aim of this paper is to give further insight into the properties of the program representation used by enzyme GP, and in particular to verify that the representation used by enzyme GP is able to support meaningful program evolution. An important issue addressed by this paper is the evolvability of different forms of representation in GP and the potential sources of evolvability within enzyme GP. For more information on evolvability in GP and biology, see Conrad (1990), Kirschner and Gerhart (1998), Altenberg (1994) and Wagner and Altenberg (1996). The paper is structured as follows. Section 2 compares approaches to program representation in GP and develops the notion of implicit context as an important mechanism for preserving the meaning of program components during evolution. Section 3 describes how implicit context is implemented in the program representation of enzyme GP. Section 4 presents experimental results and observations that support the ideas developed in Section 2; showing how implicit context in enzyme GP leads to behaviours that promote evolvability. Section 5 offers conclusions and speculates about the role of enzyme GP and evolutionary computation in understanding biological evolution.