This works if you already understand ahead of time how tuning parameters changes performance.<p>If your model is weak, then it looks like this falls back to normal Bayesian Optimization based on a Gaussian Process, which has performance of O(N^3), which means in practice you can't do more than a couple thousand samples.