PhD thesis,
Automatic Programming in an Arbitrary Language: Evolving Programs with Grammatical Evolution
University Of Limerick, Ireland, (August 2001)
Abstract
We present a novel Evolutionary Automatic Programming
system, Grammatical Evolution that is capable of
generating programs in an arbitrary language from a
binary string. Grammatical Evolution adopts a genotype
to phenotype mapping; the genotype is the raw genetic
material, analogous to the DNA of Molecular Biology,
and the phenotype the functional program that is
generated (the equivalent of proteins in Molecular
Biology). Resulting from the genotype-phenotype
distinction, and inspired by Molecular Biology, a
number of features are introduced that result in
benefits for Grammatical Evolution. We demonstrate
Grammatical Evolution's viability on a number of proof
of concept problems with performance on a par with, and
in some cases superior to Genetic Programming. An
analysis of the system is conducted in which we focus
on a number of features arising directly from the
genotype-phenotype distinction, namely the degenerate
genetic code, and the novel, wrapping operator. We
conclude the investigations with an analysis of the
effects of the genetic operator of crossover on
Grammatical Evolution, before detailing our conclusions
and outlining directions for future research.
