08.08.2013 Views

Mitchell, T. J. (2010) An exploration of evolutionary computation ...

Mitchell, T. J. (2010) An exploration of evolutionary computation ...

Mitchell, T. J. (2010) An exploration of evolutionary computation ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

purpose function optimiser by Fogel (1992). Simultaneously, Rechenberg (1965) was<br />

working independently on an adaptive optimiser known as the evolution strategy (ES).<br />

Contemporary evolution-inspired function optimisers are descended from one <strong>of</strong> these<br />

three interpretations, which have, since their introduction, been applied to an ever-<br />

increasing number <strong>of</strong> engineering problems. For a diverse list <strong>of</strong> applications the reader is<br />

directed to Schwefel and Bäck (1997), Bäck et al (1997a) and Rothlauf et al (2005).<br />

Although these three classes <strong>of</strong> EA are not without difference, each models the processes<br />

<strong>of</strong> evolution to some degree. At an abstract level, evolution can be regarded as the<br />

mechanism by which sophisticated and well adapted biological structures have come to<br />

exist: a process <strong>of</strong> natural selection which emerges when there is a superfluity <strong>of</strong> genetic<br />

material within an environment in which individuals struggle for existence.<br />

Just as a breeder chooses those individuals closest to his desired optimum, and<br />

discards the rest, so the natural environment improves the performance <strong>of</strong> a species<br />

by eliminating the less effective. Individuals possessing particular adaptations will<br />

survive better, and by virtue <strong>of</strong> the heritable nature <strong>of</strong> these adaptations, they will<br />

transmit them to their <strong>of</strong>fspring. Gradually, the adaptations will spread and<br />

improve so that the species will become better suited to the environment which it<br />

inhabits.<br />

2.2 The Evolutionary Algorithm<br />

When the principles <strong>of</strong> evolution are simulated and used to optimise solutions to<br />

Parkin (1979)<br />

engineering problems, the individuals, referred to in the above quote, are represented by a<br />

population <strong>of</strong> potential solutions. The environment, which is defined by a given objective<br />

function, quantifies the relative worth or fitness <strong>of</strong> each solution. Adaptations are<br />

introduced by recombining and mutating individuals from one generation to produce the<br />

<strong>of</strong>fspring that form the next. The elimination <strong>of</strong> less effective genetic material is facilitated<br />

by selecting those individuals with above average fitness to partake in reproduction more<br />

frequently than those with below average fitness. This selective bias introduces the notion<br />

<strong>of</strong> natural selection, enabling well adapted genes to propagate throughout subsequent<br />

generations. A simple representation <strong>of</strong> this <strong>evolutionary</strong> model is provided in figure 2.1.<br />

11

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!