Last active
April 25, 2017 15:13
-
-
Save adamwlev/19d2397ff8da1952556cf2417d965f6c to your computer and use it in GitHub Desktop.
Make a Monitor Function to Train a Gradient Booster from Sklearn with automated early stopping
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
## Makes a monitor where the mean of last x oob improvements | |
## are used to determine early stopping. This can be ammended | |
## to any stopping criteria one sees as fit - consecutive x | |
## negatives, more negatives than positives in last x, etc. | |
def make_monitor(running_mean_len): | |
def monitor(i,self,args): | |
if np.mean(self.oob_improvement_[max(0,i-running_mean_len+1):i+1])<0: | |
return True | |
else: | |
return False | |
return monitor | |
## Example use | |
from sklearn.ensemble import GradientBoostingRegressor | |
gbr = GradientBoostingRegressor(n_estimators=10000000,verbose=5) ## n_estimators can be arbitrarily high | |
monitor = make_monitor(10) ## this is a number that should be fit to a validation set | |
gbr.fit(X_train,y_train,monitor=monitor) | |
print gbr.estimators_.shape[0] | |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment