# pred_noise_bandwidth¶

• Available in: GBM

• Hyperparameter: yes

## Description¶

Use this option to specify the bandwidth (sigma) of Gaussian multiplicative noise ~N(1,sigma) for tree-node predictions. If this parameter is specified with a value greater than 0, then every leaf node prediction is randomly scaled by a number drawn from a normal distribution centered around 1 with a bandwidth given by this parameter.

This value must be >= to 0 and defaults to 0 (disabled).

## Example¶

library(h2o)
h2o.init()
# import the titanic dataset:
# This dataset is used to classify whether a passenger will survive '1' or not '0'
# original dataset can be found at https://stat.ethz.ch/R-manual/R-devel/library/datasets/html/Titanic.html
titanic <-  h2o.importFile("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/titanic.csv")

# convert response column to a factor
titanic['survived'] <- as.factor(titanic['survived'])

# set the predictor names and the response column name
# predictors include all columns except 'name' and the response column ("survived")
predictors <- setdiff(colnames(titanic), colnames(titanic)[2:3])
response <- "survived"

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

# try using the pred_noise_bandwidth parameter:
titanic_gbm <- h2o.gbm(x = predictors, y = response, training_frame = train,
validation_frame = valid,
pred_noise_bandwidth = 0.1, seed = 1234)

# print the auc for your model
print(h2o.auc(titanic_gbm, valid = TRUE))

# Example of values to grid over for pred_noise_bandwidth
# Note: this parameter is meant for much bigger datasets than the one in this example
hyper_params <- list( pred_noise_bandwidth = c(0, 0.1, 0.3) )

# 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")
grid <- h2o.grid(x = predictors, y = response, training_frame = train, validation_frame = valid,
algorithm = "gbm", grid_id = "titanic_grid", hyper_params = hyper_params,
search_criteria = list(strategy = "Cartesian"), seed = 1234)

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

import h2o
h2o.init()

# import the titanic dataset:
# This dataset is used to classify whether a passenger will survive '1' or not '0'
# original dataset can be found at https://stat.ethz.ch/R-manual/R-devel/library/datasets/html/Titanic.html
titanic = h2o.import_file("https://s3.amazonaws.com/h2o-public-test-data/smalldata/gbm_test/titanic.csv")

# convert response column to a factor
titanic['survived'] = titanic['survived'].asfactor()

# set the predictor names and the response column name
# predictors include all columns except 'name' and the response column ("survived")
predictors = titanic.columns
del predictors[1:3]
response = 'survived'

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

# try using the pred_noise_bandwidth parameter:
# initiliaze the estimator
titanic_gbm = H2OGradientBoostingEstimator(pred_noise_bandwidth = 0.1, seed = 1234)

# train the model
titanic_gbm.train(x = predictors, y = response, training_frame = train, validation_frame = valid)

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

# Example of values to grid over for pred_noise_bandwidth
# import Grid Search
from h2o.grid.grid_search import H2OGridSearch

# select the values for pred_noise_bandwidth to grid over
# Note: this parameter is meant for much bigger datasets than the one in this example
hyper_params = {'pred_noise_bandwidth': [0.0, 0.1, 0.3]}

# 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: {'strategy': "RandomDiscrete"}
# initialize the GBM estimator