Classification H2O Deep Learning Learner

Classification feed-forward multilayer artificial neural network learner. Class h2o::h2o.deeplearning() from package h2o.

Dictionary

This Learner can be instantiated via lrn():

lrn("classif.h2o.deeplearning")

Meta Information

• Task type: “classif”

• Predict Types: “response”, “prob”

• Feature Types: “integer”, “numeric”, “factor”

• Required Packages: mlr3, h2o

Parameters

Id	Type	Default	Levels	Range
ignore_const_cols	logical	TRUE	TRUE, FALSE	-
score_each_iteration	logical	FALSE	TRUE, FALSE	-
balance_classes	logical	FALSE	TRUE, FALSE	-
class_sampling_factors	untyped	NULL		-
max_after_balance_size	numeric	5		\((-\infty, \infty)\)
checkpoint	untyped	NULL		-
pretrained_autoencoder	untyped	NULL		-
overwrite_with_best_model	logical	TRUE	TRUE, FALSE	-
standardize	logical	TRUE	TRUE, FALSE	-
activation	character	Rectifier	Rectifier, Tanh, TanhWithDropout, RectifierWithDropout, Maxout, MaxoutWithDropout	-
hidden	untyped	c(200L, 200L)		-
epochs	numeric	10		\([1, \infty)\)
train_samples_per_iteration	integer	-2		\([-2, \infty)\)
target_ratio_comm_to_comp	numeric	0.05		\((-\infty, \infty)\)
seed	integer	-1		\((-\infty, \infty)\)
adaptive_rate	logical	TRUE	TRUE, FALSE	-
rho	numeric	0.99		\([0, \infty)\)
epsilon	numeric	1e-08		\([1e-10, 1e-04]\)
rate	numeric	0.005		\([0, 1]\)
rate_annealing	numeric	1e-06		\([0, \infty)\)
rate_decay	numeric	1		\([0, \infty)\)
momentum_start	numeric	0		\((-\infty, \infty)\)
momentum_ramp	numeric	1e+06		\((-\infty, \infty)\)
momentum_stable	numeric	0		\((-\infty, \infty)\)
nesterov_accelerated_gradient	logical	TRUE	TRUE, FALSE	-
input_dropout_ratio	numeric	0		\((-\infty, \infty)\)
hidden_dropout_ratios	numeric	0.5		\((-\infty, \infty)\)
l1	numeric	0		\((-\infty, \infty)\)
l2	numeric	0		\((-\infty, \infty)\)
max_w2	numeric	3.402823e+38		\((-\infty, \infty)\)
initial_weight_distribution	character	UniformAdaptive	UniformAdaptive, Uniform, Normal	-
initial_weight_scale	numeric	1		\((-\infty, \infty)\)
loss	character	Automatic	Automatic, CrossEntropy, Quadratic	-
distribution	character	AUTO	AUTO, bernoulli, multinomial	-
score_interval	numeric	5		\((-\infty, \infty)\)
score_training_samples	integer	10000		\((-\infty, \infty)\)
score_validation_samples	integer	0		\((-\infty, \infty)\)
score_duty_cycle	numeric	0.1		\((-\infty, \infty)\)
classification_stop	numeric	0		\([-1, \infty)\)
stopping_rounds	integer	5		\([0, \infty)\)
stopping_metric	character	AUTO	AUTO, logloss, AUC, lift_top_group, misclassification, AUCPR, mean_per_class_error	-
stopping_tolerance	numeric	0		\([0, \infty)\)
max_runtime_secs	numeric	0		\([0, \infty)\)
score_validation_sampling	character	Uniform	Uniform, Stratified	-
diagnostics	logical	TRUE	TRUE, FALSE	-
fast_mode	logical	TRUE	TRUE, FALSE	-
force_load_balance	logical	TRUE	TRUE, FALSE	-
replicate_training_data	logical	TRUE	TRUE, FALSE	-
single_node_mode	logical	FALSE	TRUE, FALSE	-
shuffle_training_data	logical	FALSE	TRUE, FALSE	-
missing_values_handling	character	MeanImputation	MeanImputation, Skip	-
quiet_mode	logical	TRUE	TRUE, FALSE	-
autoencoder	logical	FALSE	TRUE, FALSE	-
sparse	logical	FALSE	TRUE, FALSE	-
average_activation	numeric	0		\((-\infty, \infty)\)
sparsity_beta	numeric	0		\((-\infty, \infty)\)
max_categorical_features	integer	NULL		\((-\infty, \infty)\)
reproducible	logical	FALSE	TRUE, FALSE	-
export_weights_and_biases	logical	FALSE	TRUE, FALSE	-
mini_batch_size	integer	1		\((-\infty, \infty)\)
categorical_encoding	character	AUTO	AUTO, Enum, OneHotInternal, OneHotExplicit, Binary, Eigen, LabelEncoder, SortByResponse, EnumLimited	-
elastic_averaging	logical	FALSE	TRUE, FALSE	-
elastic_averaging_moving_rate	numeric	0.9		\((-\infty, \infty)\)
elastic_averaging_regularization	numeric	0.001		\((-\infty, \infty)\)
export_checkpoints_dir	untyped	NULL		-
auc_type	character	AUTO	AUTO, NONE, MACRO_OVR, WEIGHTED_OVR, MACRO_OVO, WEIGHTED_OVO	-
gainslift_bins	integer	-1		\([-1, \infty)\)
verbose	logical	FALSE	TRUE, FALSE	-

References

Fryda T, LeDell E, Gill N, Aiello S, Fu A, Candel A, Click C, Kraljevic T, Nykodym T, Aboyoun P, Kurka M, Malohlava M, Poirier S, Wong W (2025). h2o: R Interface for the 'H2O' Scalable Machine Learning Platform. R package version 3.46.0.9, https://github.com/h2oai/h2o-3.

See also

• Dictionary of Learners: mlr3::mlr_learners.

• as.data.table(mlr_learners) for a table of available Learners in the running session (depending on the loaded packages).

• Chapter in the mlr3book: https://mlr3book.mlr-org.com/basics.html#learners

• mlr3learners for a selection of recommended learners.

• mlr3cluster for unsupervised clustering learners.

• mlr3pipelines to combine learners with pre- and postprocessing steps.

• mlr3tuning for tuning of hyperparameters, mlr3tuningspaces for established default tuning spaces.

Author

awinterstetter

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifH2ODeeplearning

Methods

Public methods

• LearnerClassifH2ODeeplearning$new()

• LearnerClassifH2ODeeplearning$clone()

Inherited methods

• mlr3::Learner$base_learner()
• mlr3::Learner$configure()
• mlr3::Learner$encapsulate()
• mlr3::Learner$format()
• mlr3::Learner$help()
• mlr3::Learner$predict()
• mlr3::Learner$predict_newdata()
• mlr3::Learner$print()
• mlr3::Learner$reset()
• mlr3::Learner$selected_features()
• mlr3::Learner$train()
• mlr3::LearnerClassif$predict_newdata_fast()

---

Method new()

Creates a new instance of this R6 class.

Usage

LearnerClassifH2ODeeplearning$new()

---

Method clone()

The objects of this class are cloneable with this method.

Usage

LearnerClassifH2ODeeplearning$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

Examples

# Define the Learner
learner = lrn("classif.h2o.deeplearning")
print(learner)
#> 
#> ── <LearnerClassifH2ODeeplearning> (classif.h2o.deeplearning): H2O Deep Learning
#> • Model: -
#> • Parameters: list()
#> • Packages: mlr3, mlr3extralearners, and h2o
#> • Predict Types: [response] and prob
#> • Feature Types: integer, numeric, and factor
#> • Encapsulation: none (fallback: -)
#> • Properties: missings, multiclass, twoclass, and weights
#> • Other settings: use_weights = 'use'

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)
#> Warning: 
#> Your H2O cluster version is (2 years, 2 months and 9 days) old. There may be a newer version available.
#> Please download and install the latest version from: https://h2o-release.s3.amazonaws.com/h2o/latest_stable.html

print(learner$model)
#> Model Details:
#> ==============
#> 
#> H2OBinomialModel: deeplearning
#> Model ID:  DeepLearning_model_R_1772455678928_1 
#> Status of Neuron Layers: predicting Class, 2-class classification, bernoulli distribution, CrossEntropy loss, 52,802 weights/biases, 632.9 KB, 1,390 training samples, mini-batch size 1
#>   layer units      type dropout       l1       l2 mean_rate rate_rms momentum
#> 1     1    60     Input  0.00 %       NA       NA        NA       NA       NA
#> 2     2   200 Rectifier  0.00 % 0.000000 0.000000  0.006835 0.004646 0.000000
#> 3     3   200 Rectifier  0.00 % 0.000000 0.000000  0.029023 0.111292 0.000000
#> 4     4     2   Softmax      NA 0.000000 0.000000  0.002185 0.000890 0.000000
#>   mean_weight weight_rms mean_bias bias_rms
#> 1          NA         NA        NA       NA
#> 2    0.000109   0.091923  0.493855 0.009165
#> 3   -0.000741   0.069720  0.997626 0.005519
#> 4    0.001226   0.412305  0.000002 0.001697
#> 
#> 
#> H2OBinomialMetrics: deeplearning
#> ** Reported on training data. **
#> ** Metrics reported on full training frame **
#> 
#> MSE:  0.00134771
#> RMSE:  0.03671116
#> LogLoss:  0.008448052
#> Mean Per-Class Error:  0
#> AUC:  1
#> AUCPR:  1
#> Gini:  1
#> 
#> Confusion Matrix (vertical: actual; across: predicted) for F1-optimal threshold:
#>         M  R    Error    Rate
#> M      69  0 0.000000   =0/69
#> R       0 70 0.000000   =0/70
#> Totals 69 70 0.000000  =0/139
#> 
#> Maximum Metrics: Maximum metrics at their respective thresholds
#>                         metric threshold     value idx
#> 1                       max f1  0.930033  1.000000  54
#> 2                       max f2  0.930033  1.000000  54
#> 3                 max f0point5  0.930033  1.000000  54
#> 4                 max accuracy  0.930033  1.000000  54
#> 5                max precision  1.000000  1.000000   0
#> 6                   max recall  0.930033  1.000000  54
#> 7              max specificity  1.000000  1.000000   0
#> 8             max absolute_mcc  0.930033  1.000000  54
#> 9   max min_per_class_accuracy  0.930033  1.000000  54
#> 10 max mean_per_class_accuracy  0.930033  1.000000  54
#> 11                     max tns  1.000000 69.000000   0
#> 12                     max fns  1.000000 56.000000   0
#> 13                     max fps  0.000000 69.000000 123
#> 14                     max tps  0.930033 70.000000  54
#> 15                     max tnr  1.000000  1.000000   0
#> 16                     max fnr  1.000000  0.800000   0
#> 17                     max fpr  0.000000  1.000000 123
#> 18                     max tpr  0.930033  1.000000  54
#> 
#> Gains/Lift Table: Extract with `h2o.gainsLift(<model>, <data>)` or `h2o.gainsLift(<model>, valid=<T/F>, xval=<T/F>)`
#> 
#> 


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()
#> classif.ce 
#>  0.2463768