When creating a genetic algorithm, I've run paramater calibration on both small and large test instances. The best parameters for a small test case are not the best for a large test case and vice versa. I would please like some advice or recommendations on how to handle this and ensure the algorithm runs to the best of its ability.