Hi,
This is a know characteristic of Genetic Programming: the bloat effect. Some papers from Ricardo Poli have been studying that in detail. If I remember the cause of bloat effect was located in the crossover. See: "Generalisation of the limiting distribution of program sizes in tree-based genetic programming and analysis of its effects on bloat". So if you want to avoid it, control the size of the trees resulting from crossover. Don't know exactly what are the present methods in use, but I guess if you google "Bloat" and "Genetic Programming", you should get the last info on the subject. e.g. "Resource-limited genetic programming: the dynamic approach"
Good luck! |