stopping_rounds

  • Available in: GBM, DRF, Deep Learning, GLM, GAM, AutoML, XGBoost, Isolation Forest

  • Hyperparameter: yes

Description

Use this option to stop model training when the option selected for stopping_metric doesn’t improve for this specified number of training rounds, based on a simple moving average. For example, given the following options:

  • stopping_rounds=3

  • stopping_metric=misclassification

  • stopping_tolerance=1e-3

then the moving average for last 4 stopping rounds is calculated (the first moving average is reference value for other 3 moving averages to compare).

The model will stop if the ratio between the best moving average and reference moving average is more or equal 1-1e-3 (the misclassification is the less the better metric, for the more the better metrics the ratio have to be less or equal 1+1e-3 to stop).

These stopping options are used to increase performance by restricting the number of models that get built.

The default value for this option varies depending on the algorithm:

  • GBM/DRF/XGBoost: stopping_rounds defaults to 0 (disabled)

  • Deep Learning: stopping_rounds defaults to 5

  • AutoML: stopping_rounds defaults 3

To disable this feature, specify 0. When disabled, the metric is computed on the validation data (if provided); otherwise, training data is used.

When used with Deep Learning, you can also specify the overwrite_with_best_model option. When enabled, the final model is the best model generated for the given stopping_metric option.

Keep in mind that stopping_rounds does not refer to epochs, but more specifically to the number of scoring events (which can only happen after every iteration).

Notes: If cross-validation is enabled:

  • All cross-validation models stop training when the validation metric doesn’t improve.

  • The main model runs for the mean number of epochs.

  • N+1 models do not use overwrite_with_best_model, which is an available option in Deep Learning.

  • N+1 models may be off by the number specified for stopping_rounds from the best model, but the cross-validation metric estimates the performance of the main model for the resulting number of epochs (which may be fewer than the specified number of scoring events).

  • stopping_rounds must be enabled for stopping_metric or stopping_tolerance to work.

Example

R

Python

library(h2o)
h2o.init()
# import the airlines dataset:
# This dataset is used to classify whether a flight will be delayed 'YES' or not "NO"
# original data can be found at http://www.transtats.bts.gov/
airlines <-  h2o.importFile("http://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")

# convert columns to factors
airlines["Year"] <- as.factor(airlines["Year"])
airlines["Month"] <- as.factor(airlines["Month"])
airlines["DayOfWeek"] <- as.factor(airlines["DayOfWeek"])
airlines["Cancelled"] <- as.factor(airlines["Cancelled"])
airlines['FlightNum'] <- as.factor(airlines['FlightNum'])

# set the predictor names and the response column name
predictors <- c("Origin", "Dest", "Year", "UniqueCarrier", "DayOfWeek", "Month", "Distance", "FlightNum")
response <- "IsDepDelayed"

# split into train and validation
airlines_splits <- h2o.splitFrame(data =  airlines, ratios = 0.8, seed = 1234)
train <- airlines_splits[[1]]
valid <- airlines_splits[[2]]

# try using the `stopping_rounds` parameter:
# train your model, where you specify the stopping_metric, stopping_rounds,
# and stopping_tolerance
airlines_gbm <- h2o.gbm(x = predictors, y = response, training_frame = train, validation_frame = valid,
                        stopping_metric = "AUC", stopping_rounds = 3,
                        stopping_tolerance = 1e-2, seed = 1234)

# print the auc for the validation data
print(h2o.auc(airlines_gbm, valid = TRUE))

import h2o
from h2o.estimators.gbm import H2OGradientBoostingEstimator
h2o.init()
h2o.cluster().show_status()

# import the airlines dataset:
# This dataset is used to classify whether a flight will be delayed 'YES' or not "NO"
# original data can be found at http://www.transtats.bts.gov/
airlines= h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/airlines/allyears2k_headers.zip")

# convert columns to factors
airlines["Year"]= airlines["Year"].asfactor()
airlines["Month"]= airlines["Month"].asfactor()
airlines["DayOfWeek"] = airlines["DayOfWeek"].asfactor()
airlines["Cancelled"] = airlines["Cancelled"].asfactor()
airlines['FlightNum'] = airlines['FlightNum'].asfactor()

# set the predictor names and the response column name
predictors = ["Origin", "Dest", "Year", "UniqueCarrier", "DayOfWeek", "Month", "Distance", "FlightNum"]
response = "IsDepDelayed"

# split into train and validation sets
train, valid= airlines.split_frame(ratios = [.8], seed = 1234)

# try using the `stopping_rounds` parameter:
# train your model, where you specify the stopping_metric, stopping_rounds,
# and stopping_tolerance
# initialize the estimator then train the model
airlines_gbm = H2OGradientBoostingEstimator(stopping_metric = "auc", stopping_rounds = 3,
                                            stopping_tolerance = 1e-2,
                                            seed =1234)
airlines_gbm.train(x = predictors, y = response, training_frame = train, validation_frame = valid)

# print the auc for the validation data
airlines_gbm.auc(valid=True)