Ian has implemented one representation for a genetic algorithm that finds near optimal solutions for NP-complete problems e.g. Maximal Cliques, Graph Colouring, etc. Briefly, this representation when producing offspring - via traditional genetic operators such as crossover and mutation - has no illegality and therefore requires no repair algorithms. Investigations continue to see if this representation can be exploited further and produce coevolutionary model that can enforce mutualism between two or more populations each trying to find near-optimal solutions for different problem domains.
Ian has implemented a Genetic Program, GP, that instead of using the traditional S-expressions uses an XML application, MathML. Using Koza's quintic and sextic polynomials the GP successfully regressed to produced exact replicas (ignoring bloat). As expected the amount of bloat was significantly reduced when using Langdon's "Size Fair and Homologous" crossover techniques, although there are some significant differences between S-expressions and MathML that require further research. Ian is looking for funding to investigate the use of GPs to generate other XML instances e.g. XSL/T, XSD, ChemML, XHTML, etc.