Lab: Tree Ensembles

In the optional lab: Tree Ensembles, the section 4.1 suggest that we should check for some value of min_samples_split as well as value of max_depth to find the best values. Then, why don’t we check for all pairs of value? Isn’t it true that check all pair will results in the best pair, which is even better than we pick the best value from each list?

That seems computationally expensive since there can be multiple hyperparameters to tune for real-world problems, not just two.

However, if you’d like, you can experiment with these two hyperparameters in the lab.

1 Like