col_sample_rate

  • Available in: GBM, DRF
  • Hyperparameter: yes

Description

Specify the column sampling rate (y-axis). This acceptable value range is 0.0 to 1.0. Higher values may improve training accuracy. Test accuracy improves when either columns or rows are sampled. (For details, refer to “Stochastic Gradient Boosting” (Friedman, 1999)).

The following illustrates how column sampling is implemented for GBM and DRF.

For an example model using:

  • 100-column dataset
  • col_sample_rate_per_tree=0.754
  • col_sample_rate=0.8 (Samples 80% of columns per split)

For each tree, the floor is used to determine the number of columns - in this example, (0.754 * 100)=75 out of 100 - that are randomly picked, and then the floor is used to determine the number of columns - in this case, (0.754 * 0.8 * 100)=60 - that are then randomly chosen for each split decision (out of the 75).

Row and column sampling (sample_rate and col_sample_rate) can improve generalization and lead to lower validation and test set errors. Good general values for large datasets are around 0.7 to 0.8 (sampling 70-80 percent of the data) for both parameters. Column sampling per tree (col_sample_rate_per_tree) can also be used. Note that col_sample_rate_per_tree is multiplicative with col_sample_rate, so setting both parameters to 0.8, for example, results in 64% of columns being considered at any given node to split.

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 = .8, seed = 1234)
train <- airlines.splits[[1]]
valid <- airlines.splits[[2]]

# try using the `col_sample_rate` parameter:
airlines.gbm <- h2o.gbm(x = predictors, y = response, training_frame = train,
                        validation_frame = valid, col_sample_rate =.7 ,
                        seed = 1234)

# print the AUC for the validation data
print(h2o.auc(airlines.gbm, valid = TRUE))


# Example of values to grid over for `col_sample_rate`
hyper_params <- list( col_sample_rate = c(.3, .7, .8, 1) )

# this example uses cartesian grid search because the search space is small
# and we want to see the performance of all models. For a larger search space use
# random grid search instead: list(strategy = "RandomDiscrete")
# this GBM uses early stopping once the validation AUC doesn't improve by at least 0.01% for
# 5 consecutive scoring events
grid <- h2o.grid(x = predictors, y = response, training_frame = train, validation_frame = valid,
                 algorithm = "gbm", grid_id = "air_grid", hyper_params = hyper_params,
                 stopping_rounds = 5, stopping_tolerance = 1e-4, stopping_metric = "AUC",
                 search_criteria = list(strategy = "Cartesian"), seed = 1234)

## Sort the grid models by AUC
sortedGrid <- h2o.getGrid("air_grid", sort_by = "auc", decreasing = TRUE)
sortedGrid