Appendix A - ParametersΒΆ
This Appendix provides detailed descriptions of parameters that can be specified in the H2O algorithms. In addition, each parameter also includes the algorithms that support the parameter, whether the parameter is a hyperparameter (can be used in grid search), links to any related parameters, and R and Python examples showing the parameter in use.
Notes:
- This Appendix is a work in progress.
- For parameters that are supported in multiple algorithms, the included example uses the GBM or GLM algorithm.
- The examples in this section include datasets that are pulled from GitHub and S3. Alternatively, you can run the following command from within the H2O repository on your local machine to retrieve all datasets in the smalldata folder:
./gradlew syncSmalldata
alpha
balance_classes
beta_epsilon
binomial_double_trees
build_tree_one_node
categorical_encoding
checkpoint
class_sampling_factors
col_sample_rate
col_sample_rate_change_per_level
col_sample_rate_per_tree
compute_p_values
distribution
early_stopping
estimate_k
family
fold_assignment
fold_column
gradient_epsilon
histogram_type
huber_alpha
ignore_const_cols
ignored_columns
init
interactions
intercept
k
keep_cross_validation_fold_assignment
keep_cross_validation_predictions
lambda
lambda_min_ratio
lambda_search
learn_rate
learn_rate_annealing
link
max_abs_leafnode_pred
max_active_predictors
max_after_balance_size
max_depth
max_hit_ratio_k
max_iterations
max_runtime_secs
min_rows
min_split_improvement
missing_values_handling
model_id
mtries
nbins
nbins_cats
nbins_top_level
nfolds
nlambdas
non_negative
ntrees
objective_epsilon
offset_column
pred_noise_bandwidth
quantile_alpha
remove_collinear_columns
sample_rate
sample_rate_per_class
score_each_iteration
score_tree_interval
seed
solver
standardize
stopping_metric
stopping_rounds
stopping_tolerance
training_frame
tweedie_link_power
tweedie_power
tweedie_variance_power
user_points
validation_frame
weights_column
y