Towards a Rational Methodology for Using Evolutionary Search Algorithms

Evolutionary search algorithms (ESAs from now on) are iterative problem solvers developed with inspiration from neo-Darwinian survival of the fittest genes. This thesis looks at the core issues surrounding ESAs and is a step towards building a rational methodology for their effective use. Currently there is no such method of best practice rather the whole process of designing and using ESAs is seen as more of a black art than a tried and tested engineering tool. Consequently, many non-practitioners are sceptical of the worth of ESAs as a useful tool at all. Therefore the first task of the thesis is to lay out the reasons, from computational theory, why ESAs can be a potentially powerful tool. In this context the theory of NP-completeness is introduced to ground the discussions throughout the thesis. Then a generic framework for describing ESAs is developed to form another cornerstone of these later discussions. As well as arguing for their potential power, the main argument of the thesis is that no single ESA can be sufficiently powerful to fulfill all of our engineering needs. In this context the no free lunch theorem will be introduced and the arguments surrounding the notion of real world problems will be discussed in detail. While real world problems may save us from the full vigour of the no free lunch theorem, I shall argue that there is still good reason to abandon the quest for the ‘holy grail’ ESA of real world problems. Consequently our attention moves from arguing about which ESA is best to the meta-level of discussing how one could effectively go about choosing the right ESA for a given problem. With the meta-level discussion in mind a method of classifying search spaces according to their search bias characteristics is introduced. The classification technique helps discern which search biases are effective on a given search space. The technique is applied to a number of search spaces including a large real world problem to demonstrate its use. Furthermore it is argued that we need to learn more about the dynamics that results from the use of ESAs on different search problems. The dynamics of various search behaviours is examined and towards the end of the thesis a first attempt is made towards identifying which types of dynamics are those of effective search. It is argued that the holy grail of the ESA community may be a form of dynamics rather than a particular algorithm. In conclusion I sketch out the road towards a rational methodology for using ESAs, in particular I will argue that the ESA community needs to build up a corpus of well understood real world problems that have been usefully classified. However, I shall also discuss some of the major hurdles that remain in the way of a fully automated process for using evolutionary search algorithms as a tool for problem solving. Submitted for the degree of D. Phil. University of Sussex