- author = "Peter Martin",
- title = "An Investigation into the use of Genetic Programming for Intelligent Network Service Creation",
- school = "Bournemouth University",
- year = "1998",
- keywords = "genetic algorithms, genetic programming",
-
URL = "
http://www.naiadhome.com/Peter_Martin_MSC_Dissertation.pdf",
-
abstract = "Service creation is crucial to the success of
Intelligent Networks (IN). However, the time required
to develop complex services is increasing. By reducing
the elapsed time needed to generate the service logic
and by reducing the opportunity for implementation
errors to appear in the service logic, a higher quality
IN service can be delivered. This project explores an
alternative method to the existing manual service
creation, by exploiting the properties of Genetic
Programming (GP). Genetic Programming is a powerful
method for evolving computer programs via the process
of natural selection. [Koz92]. The use of Genetic
Programming to produce service logic programs for IN is
analysed and a number of key features identified.
Principally for GP to be of benefit to IN it must be
able to reduce the time to create a service and reduce
the number of implementation errors in the resultant
program. Experimental evidence is presented that shows
that using Genetic Programming is a viable method for
service creation in Intelligent Networks, and can
reduce the time to create a program by several orders
of magnitude compared to a human. The case is also
argued that since GP needs a fitness function to be
developed, the initial specification should be of a
higher quality than one produced for a human
programmer, thereby reducing the number of errors in
the final program. To implement the experimental
prototype, existing methods of evolving complex systems
using GP were researched. A new method of ensuring the
property of closure is presented that does not
constrain the development of novel service logic
implementations, in contrast to existing methods
commonly employed in GP.
Further work is identified at the end to improve upon the performance and to explore more complex services.",
-
notes = "See \cite{martin:2000:GPscin}, The project is based on
the excellent Genetic Programming Kernel by Thomas
Weinbrenner. GPSC.CC code in
http://www.naiadhome.com/gpsc.tgz
pix 'Finally I would like to thank Marconi Communications Limited, formerly GPT Limited for sponsoring me to do this MSc.'",
