My current interest in Evolutionary Algorithms is limited, but some time ago I did these:
BGP was an effort to create a fast genetic programming environment. This is done by using machine-code programs that are executed using a virtual machine, for added safety. The BGP environment was programmed in Blitz Basic and assembly language, with special optimizations for near-native VM interpretation and it features some then-state-of-the-art GP techniques like demes, and complex crossover options, but no structurul credit-assignment mechanisms. It also features informative graphical displays on the population fitness.
The process of creating an object-oriented version of BGP, for computer entertainment purposes, is underway.
My MSc thesis at the University of Essex had been involved with creating telecom networks - in particular with choosing a topology for the EON. The main difficulty was to construct a suitable function set. The function set had to be unbiased, yet efficient in searching the large problem space, something naturally done by a simple GA, but a bit more unnatural to do with GP techniqus. In my thesis, I used something similar to Luke's Edge Encoding, which was further refined my Mr. M.Sinclair after I left the Uni. E-Mail: olethros@geocities.com