Аннотация
Preface:
Since man began to dream of machines that could
automate not only the more mundane and laborious tasks
of everyday life, but that could also improve some of
the more agreeable aspects, he has turned to nature for
inspiration. This inspiration has taken all sorts of
forms, with inventors producing everything from
Icarus-like, bird-inspired flying machines to robots
based on some of the more mechanically useful human
appendages.
Another inspiration that can be taken from nature is to
employ its tools, rather than necessarily employing its
products. In this way, the field of evolutionary
computation has taken stock of the power of evolution,
and applied it, albeit at a very coarse level, to
problem solving. Genetic Programming, a powerful
incarnation of evolutionary computation uses the
artificial evolutionary process to automatically
generate programs. The adoption of evolution to
automatic generation of programs represents one of the
most promising approaches to that holy grail of
computer science, automatic programming, that is, a
computer that can automatically generate a program from
scratch given a high-level problem
description.
Research in Genetic Programming has explored a number
of program representations beyond the original Lisp
S-expression syntax trees, and some of the more
powerful of these incorporate a developmental strategy
that transforms an embryonic state into a fully fledged
adult program.
The form of Genetic Programming presented in this book,
Grammatical Evolution, delves further into nature's
processes at a molecular level, embracing the
developmental approach, and drawing upon a number of
principles that allow an abstract representation of a
program to be evolved.
This abstraction enables firstly, a separation of the
search and solution spaces that allow the EA search
engine to be a plug-in component of the system,
facilitating the exploitation of advances in EAs by GE.
Secondly, this allows the evolution of programs in an
arbitrary language with the representation of a
program's syntax in the form of a grammar
definition.
Thirdly, the existence of a degenerate genetic code is
enabled, giving a many-to-one mapping, that allows the
exploitation of neutral evolution to enhance the search
efficiency of the EA. Fourthly, we can adopt the use of
a wrapping operator that allows the reuse of genetic
material during a genotype-phenotype mapping
process.
This book is partly based on the Ph.D. thesis of
Michael O'Neill, and reports a number of new directions
in Grammatical Evolution research that are been
conducted both within the confines of the University of
Limerick's Biocomputing-Developmental Systems Centre
where the book's authors reside, and also developments
that are occurring through collaborations around the
globe.
Линки и ресурсы
тэги