From: Klaus Krogmann <krogmann@ip...> - 2007-08-08 09:28:54
is there an intended way to improve the convergence speed in GPProgramming by providing a kind of initially "predefined" chromosome? I already tried using a prototype program and an GPInitStrategy. As far as I checked the code prototype programs are currently used as a fallback only if the creation of a random GPProgram fails (number of tries exceeds maxTries from config) and GPInitStrategies are returning single CommandGenes only.
It would be sufficient for me if I could provide a set of GPPrograms or ProgramChromosomes for the first generation.