Convergence

Module 6. Ensemble of models Wrap-up quiz
#1

Excuse me, I set max_iter (maximum number of trees in HistGradientBoostingClassifier) but If I need to know the numbers of trees to convergence, how can I obtain that number?
Thanks

#2

You can force early_stopping=True and put a large number of iteration. In this case, the training will stop when the convergence requirements are met. These requirements are based on tol and n_iter_no_change. To know the size of the example once trained, you can check the fitted attribute model.n_iter_.

#3

Thank you, I have launched this code:

from sklearn.experimental import enable_hist_gradient_boosting
from sklearn.ensemble import HistGradientBoostingClassifier

hist_gbdt = HistGradientBoostingClassifier(
    max_iter=1000, early_stopping=True, random_state=0
)
cv_results = cross_validate(
    hist_gbdt, data, target, cv=10, scoring="balanced_accuracy", n_jobs=2,
    return_estimator=True,
)
cv_results["test_score"].mean()

and the when I need to know the number on average of trees to convergence I do not cath the correct sentence, can you help me, please?

#4

So you need to access to the fitted estimator stored in the cv_results. For instance, to access to the estimator fitted on the first CV iteration, you can do:

cv_results["estimator"][0].n_iter_

So you can check for all CV iterations and average them.