Modeling In H2O¶

Bases: h2o.estimators.estimator_base.H2OEstimator

Deep Learning

Build a Deep Neural Network model using CPUs Builds a feed-forward multilayer artificial neural network on an H2OFrame

Examples

>>> import h2o
>>> from h2o.estimators.deeplearning import H2ODeepLearningEstimator
>>> h2o.connect()
>>> rows = [[1,2,3,4,0], [2,1,2,4,1], [2,1,4,2,1], [0,1,2,34,1], [2,3,4,1,0]] * 50
>>> fr = h2o.H2OFrame(rows)
>>> fr[4] = fr[4].asfactor()
>>> model = H2ODeepLearningEstimator()
>>> model.train(x=range(4), y=4, training_frame=fr)

activation¶

Activation function.

One of: "tanh", "tanh_with_dropout", "rectifier", "rectifier_with_dropout", "maxout", "maxout_with_dropout" (default: "rectifier").

adaptive_rate¶

Adaptive learning rate.

Type: bool (default: True).

autoencoder¶

Auto-Encoder.

Type: bool (default: False).

average_activation¶

Average activation for sparse auto-encoder. #Experimental

Type: float (default: 0).

balance_classes¶

Balance training data class counts via over/under-sampling (for imbalanced data).

Type: bool (default: False).

categorical_encoding¶

Encoding scheme for categorical features

One of: "auto", "enum", "one_hot_internal", "one_hot_explicit", "binary", "eigen" (default: "auto").

checkpoint¶

Model checkpoint to resume training with.

Type: str.

class_sampling_factors¶

Desired over/under-sampling ratios per class (in lexicographic order). If not specified, sampling factors will be automatically computed to obtain class balance during training. Requires balance_classes.

Type: List[float].

classification_stop¶

Stopping criterion for classification error fraction on training data (-1 to disable).

Type: float (default: 0).

col_major¶

#DEPRECATED Use a column major weight matrix for input layer. Can speed up forward propagation, but might slow down backpropagation.

Type: bool (default: False).

diagnostics¶

Enable diagnostics for hidden layers.

Type: bool (default: True).

distribution¶

Distribution function

One of: "auto", "bernoulli", "multinomial", "gaussian", "poisson", "gamma", "tweedie", "laplace", "quantile", "huber" (default: "auto").

elastic_averaging¶

Elastic averaging between compute nodes can improve distributed model convergence. #Experimental

Type: bool (default: False).

elastic_averaging_moving_rate¶

Elastic averaging moving rate (only if elastic averaging is enabled).

Type: float (default: 0.9).

elastic_averaging_regularization¶

Elastic averaging regularization strength (only if elastic averaging is enabled).

Type: float (default: 0.001).

epochs¶

How many times the dataset should be iterated (streamed), can be fractional.

Type: float (default: 10).

epsilon¶

Adaptive learning rate smoothing factor (to avoid divisions by zero and allow progress).

Type: float (default: 1e-08).

export_weights_and_biases¶

Whether to export Neural Network weights and biases to H2O Frames.

Type: bool (default: False).

fast_mode¶

Enable fast mode (minor approximation in back-propagation).

Type: bool (default: True).

fold_assignment¶

Cross-validation fold assignment scheme, if fold_column is not specified. The ‘Stratified’ option will stratify the folds based on the response variable, for classification problems.

One of: "auto", "random", "modulo", "stratified" (default: "auto").

fold_column¶

Column with cross-validation fold index assignment per observation.

Type: str.

force_load_balance¶

Force extra load balancing to increase training speed for small datasets (to keep all cores busy).

Type: bool (default: True).

hidden¶

Hidden layer sizes (e.g. [100, 100]).

Type: List[int] (default: [200, 200]).

hidden_dropout_ratios¶

Hidden layer dropout ratios (can improve generalization), specify one value per hidden layer, defaults to 0.5.

Type: List[float].

huber_alpha¶

Desired quantile for Huber/M-regression (threshold between quadratic and linear loss, must be between 0 and 1).

Type: float (default: 0.9).

ignore_const_cols¶

Ignore constant columns.

Type: bool (default: True).

ignored_columns¶

Names of columns to ignore for training.

Type: List[str].

initial_biases¶

A list of H2OFrame ids to initialize the bias vectors of this model with.

Type: List[str].

initial_weight_distribution¶

Initial weight distribution.

One of: "uniform_adaptive", "uniform", "normal" (default: "uniform_adaptive").

initial_weight_scale¶

Uniform: -value...value, Normal: stddev.

Type: float (default: 1).

initial_weights¶

A list of H2OFrame ids to initialize the weight matrices of this model with.

Type: List[str].

input_dropout_ratio¶

Input layer dropout ratio (can improve generalization, try 0.1 or 0.2).

Type: float (default: 0).

keep_cross_validation_fold_assignment¶

Whether to keep the cross-validation fold assignment.

Type: bool (default: False).

keep_cross_validation_predictions¶

Whether to keep the predictions of the cross-validation models.

Type: bool (default: False).

l1¶

L1 regularization (can add stability and improve generalization, causes many weights to become 0).

Type: float (default: 0).

l2¶

L2 regularization (can add stability and improve generalization, causes many weights to be small.

Type: float (default: 0).

loss¶

Loss function.

One of: "automatic", "cross_entropy", "quadratic", "huber", "absolute", "quantile" (default: "automatic").

max_after_balance_size¶

Maximum relative size of the training data after balancing class counts (can be less than 1.0). Requires balance_classes.

Type: float (default: 5).

max_categorical_features¶

Max. number of categorical features, enforced via hashing. #Experimental

Type: int (default: 2147483647).

max_confusion_matrix_size¶

[Deprecated] Maximum size (# classes) for confusion matrices to be printed in the Logs.

Type: int (default: 20).

max_hit_ratio_k¶

Max. number (top K) of predictions to use for hit ratio computation (for multi-class only, 0 to disable).

Type: int (default: 0).

max_runtime_secs¶

Maximum allowed runtime in seconds for model training. Use 0 to disable.

Type: float (default: 0).

max_w2¶

Constraint for squared sum of incoming weights per unit (e.g. for Rectifier).

Type: float (default: 3.4028235e+38).

mini_batch_size¶

Mini-batch size (smaller leads to better fit, larger can speed up and generalize better).

Type: int (default: 1).

missing_values_handling¶

Handling of missing values. Either MeanImputation or Skip.

One of: "mean_imputation", "skip" (default: "mean_imputation").

momentum_ramp¶

Number of training samples for which momentum increases.

Type: float (default: 1000000).

momentum_stable¶

Final momentum after the ramp is over (try 0.99).

Type: float (default: 0).

momentum_start¶

Initial momentum at the beginning of training (try 0.5).

Type: float (default: 0).

nesterov_accelerated_gradient¶

Use Nesterov accelerated gradient (recommended).

Type: bool (default: True).

nfolds¶

Number of folds for N-fold cross-validation (0 to disable or >= 2).

Type: int (default: 0).

offset_column¶

Offset column. This will be added to the combination of columns before applying the link function.

Type: str.

overwrite_with_best_model¶

If enabled, override the final model with the best model found during training.

Type: bool (default: True).

pretrained_autoencoder¶

Pretrained autoencoder model to initialize this model with.

Type: str.

quantile_alpha¶

Desired quantile for Quantile regression, must be between 0 and 1.

Type: float (default: 0.5).

quiet_mode¶

Enable quiet mode for less output to standard output.

Type: bool (default: False).

rate¶

Learning rate (higher => less stable, lower => slower convergence).

Type: float (default: 0.005).

rate_annealing¶

Learning rate annealing: rate / (1 + rate_annealing * samples).

Type: float (default: 1e-06).

rate_decay¶

Learning rate decay factor between layers (N-th layer: rate * rate_decay ^ (n - 1).

Type: float (default: 1).

regression_stop¶

Stopping criterion for regression error (MSE) on training data (-1 to disable).

Type: float (default: 1e-06).

replicate_training_data¶

Replicate the entire training dataset onto every node for faster training on small datasets.

Type: bool (default: True).

reproducible¶

Force reproducibility on small data (will be slow - only uses 1 thread).

Type: bool (default: False).

response_column¶

Response variable column.

Type: str.

rho¶

Adaptive learning rate time decay factor (similarity to prior updates).

Type: float (default: 0.99).

score_duty_cycle¶

Maximum duty cycle fraction for scoring (lower: more training, higher: more scoring).

Type: float (default: 0.1).

score_each_iteration¶

Whether to score during each iteration of model training.

Type: bool (default: False).

score_interval¶

Shortest time interval (in seconds) between model scoring.

Type: float (default: 5).

score_training_samples¶

Number of training set samples for scoring (0 for all).

Type: int (default: 10000).

score_validation_samples¶

Number of validation set samples for scoring (0 for all).

Type: int (default: 0).

score_validation_sampling¶

Method used to sample validation dataset for scoring.

One of: "uniform", "stratified" (default: "uniform").

seed¶

Seed for random numbers (affects sampling) - Note: only reproducible when running single threaded.

Type: int (default: -1).

shuffle_training_data¶

Enable shuffling of training data (recommended if training data is replicated and train_samples_per_iteration is close to #nodes x #rows, of if using balance_classes).

Type: bool (default: False).

single_node_mode¶

Run on a single node for fine-tuning of model parameters.

Type: bool (default: False).

sparse¶

Sparse data handling (more efficient for data with lots of 0 values).

Type: bool (default: False).

sparsity_beta¶

Sparsity regularization. #Experimental

Type: float (default: 0).

standardize¶

If enabled, automatically standardize the data. If disabled, the user must provide properly scaled input data.

Type: bool (default: True).

stopping_metric¶

Metric to use for early stopping (AUTO: logloss for classification, deviance for regression)

One of: "auto", "deviance", "logloss", "mse", "rmse", "mae", "rmsle", "auc", "lift_top_group", "misclassification", "mean_per_class_error" (default: "auto").

stopping_rounds¶

Early stopping based on convergence of stopping_metric. Stop if simple moving average of length k of the stopping_metric does not improve for k:=stopping_rounds scoring events (0 to disable)

Type: int (default: 5).

stopping_tolerance¶

Relative tolerance for metric-based stopping criterion (stop if relative improvement is not at least this much)

Type: float (default: 0).

target_ratio_comm_to_comp¶

Target ratio of communication overhead to computation. Only for multi-node operation and train_samples_per_iteration = -2 (auto-tuning).

Type: float (default: 0.05).

train_samples_per_iteration¶

Number of training samples (globally) per MapReduce iteration. Special values are 0: one epoch, -1: all available data (e.g., replicated training data), -2: automatic.

Type: int (default: -2).

training_frame¶

Id of the training data frame (Not required, to allow initial validation of model parameters).

Type: str.

tweedie_power¶

Tweedie power for Tweedie regression, must be between 1 and 2.

Type: float (default: 1.5).

use_all_factor_levels¶

Use all factor levels of categorical variables. Otherwise, the first factor level is omitted (without loss of accuracy). Useful for variable importances and auto-enabled for autoencoder.

Type: bool (default: True).

validation_frame¶

Id of the validation data frame.

Type: str.

variable_importances¶

Compute variable importances for input features (Gedeon method) - can be slow for large networks.

Type: bool (default: False).

weights_column¶

Column with observation weights. Giving some observation a weight of zero is equivalent to excluding it from the dataset; giving an observation a relative weight of 2 is equivalent to repeating that row twice. Negative weights are not allowed.

Type: str.

Bases: h2o.estimators.estimator_base.H2OEstimator

Deep Water

Build a Deep Learning model using multiple native GPU backends Builds a deep neural network on an H2OFrame containing various data sources

activation¶

Activation function. Only used if no user-defined network architecture file is provided, and only for problem_type=dataset.

One of: "rectifier", "tanh".

autoencoder¶

Auto-Encoder.

Type: bool (default: False).

static available()[source]¶

Returns True if a deep water model can be built, or False otherwise.

backend¶

Deep Learning Backend.

One of: "mxnet", "caffe", "tensorflow" (default: "mxnet").

balance_classes¶

Balance training data class counts via over/under-sampling (for imbalanced data).

Type: bool (default: False).

categorical_encoding¶

Encoding scheme for categorical features

One of: "auto", "enum", "one_hot_internal", "one_hot_explicit", "binary", "eigen" (default: "auto").

channels¶

Number of (color) channels.

Type: int (default: 3).

checkpoint¶

Model checkpoint to resume training with.

Type: str.

class_sampling_factors¶

Desired over/under-sampling ratios per class (in lexicographic order). If not specified, sampling factors will be automatically computed to obtain class balance during training. Requires balance_classes.

Type: List[float].

classification_stop¶

Stopping criterion for classification error fraction on training data (-1 to disable).

Type: float (default: 0).

clip_gradient¶

Clip gradients once their absolute value is larger than this value.

Type: float (default: 10).

device_id¶

Device IDs (which GPUs to use).

Type: List[int] (default: [0]).

distribution¶

Distribution function

One of: "auto", "bernoulli", "multinomial", "gaussian", "poisson", "gamma", "tweedie", "laplace", "quantile", "huber" (default: "auto").

epochs¶

How many times the dataset should be iterated (streamed), can be fractional.

Type: float (default: 10).

export_native_parameters_prefix¶

Path (prefix) where to export the native model parameters after every iteration.

Type: str.

fold_assignment¶

Cross-validation fold assignment scheme, if fold_column is not specified. The ‘Stratified’ option will stratify the folds based on the response variable, for classification problems.

One of: "auto", "random", "modulo", "stratified" (default: "auto").

fold_column¶

Column with cross-validation fold index assignment per observation.

Type: str.

gpu¶

Whether to use a GPU (if available).

Type: bool (default: True).

hidden¶

Hidden layer sizes (e.g. [200, 200]). Only used if no user-defined network architecture file is provided, and only for problem_type=dataset.

Type: List[int].

hidden_dropout_ratios¶

Hidden layer dropout ratios (can improve generalization), specify one value per hidden layer, defaults to 0.5.

Type: List[float].

ignore_const_cols¶

Ignore constant columns.

Type: bool (default: True).

ignored_columns¶

Names of columns to ignore for training.

Type: List[str].

image_shape¶

Width and height of image.

Type: List[int] (default: [0, 0]).

input_dropout_ratio¶

Input layer dropout ratio (can improve generalization, try 0.1 or 0.2).

Type: float (default: 0).

keep_cross_validation_fold_assignment¶

Whether to keep the cross-validation fold assignment.

Type: bool (default: False).

keep_cross_validation_predictions¶

Whether to keep the predictions of the cross-validation models.

Type: bool (default: False).

learning_rate¶

Learning rate (higher => less stable, lower => slower convergence).

Type: float (default: 0.005).

learning_rate_annealing¶

Learning rate annealing: rate / (1 + rate_annealing * samples).

Type: float (default: 1e-06).

max_after_balance_size¶

Maximum relative size of the training data after balancing class counts (can be less than 1.0). Requires balance_classes.

Type: float (default: 5).

max_runtime_secs¶

Maximum allowed runtime in seconds for model training. Use 0 to disable.

Type: float (default: 0).

mean_image_file¶

Path of file containing the mean image data for data normalization.

Type: str.

mini_batch_size¶

Mini-batch size (smaller leads to better fit, larger can speed up and generalize better).

Type: int (default: 32).

momentum_ramp¶

Number of training samples for which momentum increases.

Type: float (default: 10000).

momentum_stable¶

Final momentum after the ramp is over (try 0.99).

Type: float (default: 0.99).

momentum_start¶

Initial momentum at the beginning of training (try 0.5).

Type: float (default: 0.9).

network¶

Network architecture.

One of: "auto", "user", "lenet", "alexnet", "vgg", "googlenet", "inception_bn", "resnet" (default: "auto").

network_definition_file¶

Path of file containing network definition (graph, architecture).

Type: str.

network_parameters_file¶

Path of file containing network (initial) parameters (weights, biases).

Type: str.

nfolds¶

Number of folds for N-fold cross-validation (0 to disable or >= 2).

Type: int (default: 0).

offset_column¶

Offset column. This will be added to the combination of columns before applying the link function.

Type: str.

overwrite_with_best_model¶

If enabled, override the final model with the best model found during training.

Type: bool (default: True).

problem_type¶

Problem type, auto-detected by default. If set to image, the H2OFrame must contain a string column containing the path (URI or URL) to the images in the first column. If set to text, the H2OFrame must contain a string column containing the text in the first column. If set to dataset, Deep Water behaves just like any other H2O Model and builds a model on the provided H2OFrame (non-String columns).

One of: "auto", "image", "dataset" (default: "auto").

regression_stop¶

Stopping criterion for regression error (MSE) on training data (-1 to disable).

Type: float (default: 0).

response_column¶

Response variable column.

Type: str.

score_duty_cycle¶

Maximum duty cycle fraction for scoring (lower: more training, higher: more scoring).

Type: float (default: 0.1).

score_each_iteration¶

Whether to score during each iteration of model training.

Type: bool (default: False).

score_interval¶

Shortest time interval (in seconds) between model scoring.

Type: float (default: 5).

score_training_samples¶

Number of training set samples for scoring (0 for all).

Type: int (default: 10000).

score_validation_samples¶

Number of validation set samples for scoring (0 for all).

Type: int (default: 0).

seed¶

Seed for random numbers (affects sampling) - Note: only reproducible when running single threaded.

Type: int (default: -1).

shuffle_training_data¶

Enable global shuffling of training data.

Type: bool (default: True).

sparse¶

Sparse data handling (more efficient for data with lots of 0 values).

Type: bool (default: False).

standardize¶

If enabled, automatically standardize the data. If disabled, the user must provide properly scaled input data.

Type: bool (default: True).

stopping_metric¶

Metric to use for early stopping (AUTO: logloss for classification, deviance for regression)

One of: "auto", "deviance", "logloss", "mse", "rmse", "mae", "rmsle", "auc", "lift_top_group", "misclassification", "mean_per_class_error" (default: "auto").

stopping_rounds¶

Early stopping based on convergence of stopping_metric. Stop if simple moving average of length k of the stopping_metric does not improve for k:=stopping_rounds scoring events (0 to disable)

Type: int (default: 5).

stopping_tolerance¶

Relative tolerance for metric-based stopping criterion (stop if relative improvement is not at least this much)

Type: float (default: 0).

target_ratio_comm_to_comp¶

Target ratio of communication overhead to computation. Only for multi-node operation and train_samples_per_iteration = -2 (auto-tuning).

Type: float (default: 0.05).

train_samples_per_iteration¶

Number of training samples (globally) per MapReduce iteration. Special values are 0: one epoch, -1: all available data (e.g., replicated training data), -2: automatic.

Type: int (default: -2).

training_frame¶

Id of the training data frame (Not required, to allow initial validation of model parameters).

Type: str.

validation_frame¶

Id of the validation data frame.

Type: str.

weights_column¶

Column with observation weights. Giving some observation a weight of zero is equivalent to excluding it from the dataset; giving an observation a relative weight of 2 is equivalent to repeating that row twice. Negative weights are not allowed.

Type: str.

Bases: h2o.estimators.estimator_base.H2OEstimator

Gradient Boosting Machine

Builds gradient boosted trees on a parsed data set, for regression or classification. The default distribution function will guess the model type based on the response column type. Otherwise, the response column must be an enum for “bernoulli” or “multinomial”, and numeric for all other distributions.

balance_classes¶

Balance training data class counts via over/under-sampling (for imbalanced data).

Type: bool (default: False).

build_tree_one_node¶

Run on one node only; no network overhead but fewer cpus used. Suitable for small datasets.

Type: bool (default: False).

categorical_encoding¶

Encoding scheme for categorical features

One of: "auto", "enum", "one_hot_internal", "one_hot_explicit", "binary", "eigen" (default: "auto").

checkpoint¶

Model checkpoint to resume training with.

Type: str.

class_sampling_factors¶

Desired over/under-sampling ratios per class (in lexicographic order). If not specified, sampling factors will be automatically computed to obtain class balance during training. Requires balance_classes.

Type: List[float].

col_sample_rate¶

Column sample rate (from 0.0 to 1.0)

Type: float (default: 1).

col_sample_rate_change_per_level¶

Relative change of the column sampling rate for every level (from 0.0 to 2.0)

Type: float (default: 1).

col_sample_rate_per_tree¶

Column sample rate per tree (from 0.0 to 1.0)

Type: float (default: 1).

distribution¶

Distribution function

One of: "auto", "bernoulli", "multinomial", "gaussian", "poisson", "gamma", "tweedie", "laplace", "quantile", "huber" (default: "auto").

fold_assignment¶

Cross-validation fold assignment scheme, if fold_column is not specified. The ‘Stratified’ option will stratify the folds based on the response variable, for classification problems.

One of: "auto", "random", "modulo", "stratified" (default: "auto").

fold_column¶

Column with cross-validation fold index assignment per observation.

Type: str.

histogram_type¶

What type of histogram to use for finding optimal split points

One of: "auto", "uniform_adaptive", "random", "quantiles_global", "round_robin" (default: "auto").

huber_alpha¶

Desired quantile for Huber/M-regression (threshold between quadratic and linear loss, must be between 0 and 1).

Type: float (default: 0.9).

ignore_const_cols¶

Ignore constant columns.

Type: bool (default: True).

ignored_columns¶

Names of columns to ignore for training.

Type: List[str].

keep_cross_validation_fold_assignment¶

Whether to keep the cross-validation fold assignment.

Type: bool (default: False).

keep_cross_validation_predictions¶

Whether to keep the predictions of the cross-validation models.

Type: bool (default: False).

learn_rate¶

Learning rate (from 0.0 to 1.0)

Type: float (default: 0.1).

learn_rate_annealing¶

Scale the learning rate by this factor after each tree (e.g., 0.99 or 0.999)

Type: float (default: 1).

max_abs_leafnode_pred¶

Maximum absolute value of a leaf node prediction

Type: float (default: 1.797693135e+308).

max_after_balance_size¶

Maximum relative size of the training data after balancing class counts (can be less than 1.0). Requires balance_classes.

Type: float (default: 5).

max_confusion_matrix_size¶

[Deprecated] Maximum size (# classes) for confusion matrices to be printed in the Logs

Type: int (default: 20).

max_depth¶

Maximum tree depth.

Type: int (default: 5).

max_hit_ratio_k¶

Max. number (top K) of predictions to use for hit ratio computation (for multi-class only, 0 to disable)

Type: int (default: 0).

max_runtime_secs¶

Maximum allowed runtime in seconds for model training. Use 0 to disable.

Type: float (default: 0).

min_rows¶

Fewest allowed (weighted) observations in a leaf.

Type: float (default: 10).

min_split_improvement¶

Minimum relative improvement in squared error reduction for a split to happen

Type: float (default: 1e-05).

nbins¶

For numerical columns (real/int), build a histogram of (at least) this many bins, then split at the best point

Type: int (default: 20).

nbins_cats¶

For categorical columns (factors), build a histogram of this many bins, then split at the best point. Higher values can lead to more overfitting.

Type: int (default: 1024).

nbins_top_level¶

For numerical columns (real/int), build a histogram of (at most) this many bins at the root level, then decrease by factor of two per level

Type: int (default: 1024).

nfolds¶

Number of folds for N-fold cross-validation (0 to disable or >= 2).

Type: int (default: 0).

ntrees¶

Number of trees.

Type: int (default: 50).

offset_column¶

Offset column. This will be added to the combination of columns before applying the link function.

Type: str.

pred_noise_bandwidth¶

Bandwidth (sigma) of Gaussian multiplicative noise ~N(1,sigma) for tree node predictions

Type: float (default: 0).

quantile_alpha¶

Desired quantile for Quantile regression, must be between 0 and 1.

Type: float (default: 0.5).

r2_stopping¶

r2_stopping is no longer supported and will be ignored if set - please use stopping_rounds, stopping_metric and stopping_tolerance instead. Previous version of H2O would stop making trees when the R^2 metric equals or exceeds this

Type: float (default: 1.797693135e+308).

response_column¶

Response variable column.

Type: str.

sample_rate¶

Row sample rate per tree (from 0.0 to 1.0)

Type: float (default: 1).

sample_rate_per_class¶

Row sample rate per tree per class (from 0.0 to 1.0)

Type: List[float].

score_each_iteration¶

Whether to score during each iteration of model training.

Type: bool (default: False).

score_tree_interval¶

Score the model after every so many trees. Disabled if set to 0.

Type: int (default: 0).

seed¶

Seed for pseudo random number generator (if applicable)

Type: int (default: -1).

stopping_metric¶

Metric to use for early stopping (AUTO: logloss for classification, deviance for regression)

One of: "auto", "deviance", "logloss", "mse", "rmse", "mae", "rmsle", "auc", "lift_top_group", "misclassification", "mean_per_class_error" (default: "auto").

stopping_rounds¶

Early stopping based on convergence of stopping_metric. Stop if simple moving average of length k of the stopping_metric does not improve for k:=stopping_rounds scoring events (0 to disable)

Type: int (default: 0).

stopping_tolerance¶

Relative tolerance for metric-based stopping criterion (stop if relative improvement is not at least this much)

Type: float (default: 0.001).

training_frame¶

Id of the training data frame (Not required, to allow initial validation of model parameters).

Type: str.

tweedie_power¶

Tweedie power for Tweedie regression, must be between 1 and 2.

Type: float (default: 1.5).

validation_frame¶

Id of the validation data frame.

Type: str.

weights_column¶

Column with observation weights. Giving some observation a weight of zero is equivalent to excluding it from the dataset; giving an observation a relative weight of 2 is equivalent to repeating that row twice. Negative weights are not allowed.

Type: str.

Bases: h2o.estimators.estimator_base.H2OEstimator

Generalized Linear Modeling

Fits a generalized linear model, specified by a response variable, a set of predictors, and a description of the error distribution.

A subclass of ModelBase is returned. The specific subclass depends on the machine learning task at hand (if it’s binomial classification, then an H2OBinomialModel is returned, if it’s regression then a H2ORegressionModel is returned). The default print-out of the models is shown, but further GLM-specific information can be queried out of the object. Upon completion of the GLM, the resulting object has coefficients, normalized coefficients, residual/null deviance, aic, and a host of model metrics including MSE, AUC (for logistic regression), degrees of freedom, and confusion matrices.

Lambda¶

DEPRECATED. Use self.lambda_ instead

alpha¶

distribution of regularization between L1 and L2. Default value of alpha is 0 when SOLVER = ‘L-BFGS’, 0.5 otherwise

Type: List[float].

balance_classes¶

Balance training data class counts via over/under-sampling (for imbalanced data).

Type: bool (default: False).

beta_constraints¶

beta constraints

Type: str.

beta_epsilon¶

converge if beta changes less (using L-infinity norm) than beta esilon, ONLY applies to IRLSM solver

Type: float (default: 0.0001).

class_sampling_factors¶

Desired over/under-sampling ratios per class (in lexicographic order). If not specified, sampling factors will be automatically computed to obtain class balance during training. Requires balance_classes.

Type: List[float].

compute_p_values¶

request p-values computation, p-values work only with IRLSM solver and no regularization

Type: bool (default: False).

early_stopping¶

stop early when there is no more relative improvement on train or validation (if provided)

Type: bool (default: True).

family¶

Family. Use binomial for classification with logistic regression, others are for regression problems.

One of: "gaussian", "binomial", "quasibinomial", "multinomial", "poisson", "gamma", "tweedie" (default: "gaussian").

fold_assignment¶

Cross-validation fold assignment scheme, if fold_column is not specified. The ‘Stratified’ option will stratify the folds based on the response variable, for classification problems.

One of: "auto", "random", "modulo", "stratified" (default: "auto").

fold_column¶

Column with cross-validation fold index assignment per observation.

Type: str.

static getGLMRegularizationPath(model)[source]¶

Extract full regularization path explored during lambda search from glm model.

Parameters:	model – source lambda search model

gradient_epsilon¶

Converge if objective changes less (using L-infinity norm) than this, ONLY applies to L-BFGS solver. Default indicates: If lambda_search is set to False and lambda is equal to zero, the default value of gradient_epsilon is equal to .000001, otherwise the default value is .0001. If lambda_search is set to True, the conditional values above are 1E-8 and 1E-6 respectively.

Type: float (default: -1).

ignore_const_cols¶

Ignore constant columns.

Type: bool (default: True).

ignored_columns¶

Names of columns to ignore for training.

Type: List[str].

interactions¶

A list of predictor column indices to interact. All pairwise combinations will be computed for the list.

Type: List[str].

intercept¶

include constant term in the model

Type: bool (default: True).

keep_cross_validation_fold_assignment¶

Whether to keep the cross-validation fold assignment.

Type: bool (default: False).

keep_cross_validation_predictions¶

Whether to keep the predictions of the cross-validation models.

Type: bool (default: False).

lambda_¶

regularization strength

Type: List[float].

lambda_min_ratio¶

Min lambda used in lambda search, specified as a ratio of lambda_max. Default indicates: if the number of observations is greater than the number of variables then lambda_min_ratio is set to 0.0001; if the number of observations is less than the number of variables then lambda_min_ratio is set to 0.01.

Type: float (default: -1).

lambda_search¶

use lambda search starting at lambda max, given lambda is then interpreted as lambda min

Type: bool (default: False).

link¶

One of: "family_default", "identity", "logit", "log", "inverse", "tweedie" (default: "family_default").

static makeGLMModel(model, coefs, threshold=0.5)[source]¶

Create a custom GLM model using the given coefficients.

Needs to be passed source model trained on the dataset to extract the dataset information from.

Parameters:	model – source model, used for extracting dataset information coefs – dictionary containing model coefficients threshold – (optional, only for binomial) decision threshold used for classification

max_active_predictors¶

Maximum number of active predictors during computation. Use as a stopping criterion to prevent expensive model building with many predictors. Default indicates: If the IRLSM solver is used, the value of max_active_predictors is set to 7000 otherwise it is set to 100000000.

Type: int (default: -1).

max_after_balance_size¶

Maximum relative size of the training data after balancing class counts (can be less than 1.0). Requires balance_classes.

Type: float (default: 5).

max_confusion_matrix_size¶

[Deprecated] Maximum size (# classes) for confusion matrices to be printed in the Logs

Type: int (default: 20).

max_hit_ratio_k¶

Max. number (top K) of predictions to use for hit ratio computation (for multi-class only, 0 to disable)

Type: int (default: 0).

max_iterations¶

Maximum number of iterations

Type: int (default: -1).

max_runtime_secs¶

Maximum allowed runtime in seconds for model training. Use 0 to disable.

Type: float (default: 0).

missing_values_handling¶

Handling of missing values. Either MeanImputation or Skip.

One of: "mean_imputation", "skip" (default: "mean_imputation").

nfolds¶

Number of folds for N-fold cross-validation (0 to disable or >= 2).

Type: int (default: 0).

nlambdas¶

Number of lambdas to be used in a search. Default indicates: If alpha is zero, with lambda search set to True, the value of nlamdas is set to 30 (fewer lambdas are needed for ridge regression) otherwise it is set to 100.

Type: int (default: -1).

non_negative¶

Restrict coefficients (not intercept) to be non-negative

Type: bool (default: False).

objective_epsilon¶

Converge if objective value changes less than this. Default indicates: If lambda_search is set to True the value of objective_epsilon is set to .0001. If the lambda_search is set to False and lambda is equal to zero, the value of objective_epsilon is set to .000001, for any other value of lambda the default value of objective_epsilon is set to .0001.

Type: float (default: -1).

offset_column¶

Offset column. This will be added to the combination of columns before applying the link function.

Type: str.

prior¶

prior probability for y==1. To be used only for logistic regression iff the data has been sampled and the mean of response does not reflect reality.

Type: float (default: -1).

remove_collinear_columns¶

in case of linearly dependent columns remove some of the dependent columns

Type: bool (default: False).

response_column¶

Response variable column.

Type: str.

score_each_iteration¶

Whether to score during each iteration of model training.

Type: bool (default: False).

seed¶

Seed for pseudo random number generator (if applicable)

Type: int (default: -1).

solver¶

AUTO will set the solver based on given data and the other parameters. IRLSM is fast on on problems with small number of predictors and for lambda-search with L1 penalty, L_BFGS scales better for datasets with many columns. Coordinate descent is experimental (beta).

One of: "auto", "irlsm", "l_bfgs", "coordinate_descent_naive", "coordinate_descent" (default: "auto").

standardize¶

Standardize numeric columns to have zero mean and unit variance

Type: bool (default: True).

training_frame¶

Id of the training data frame (Not required, to allow initial validation of model parameters).

Type: str.

tweedie_link_power¶

Tweedie link power

Type: float (default: 1).

tweedie_variance_power¶

Tweedie variance power

Type: float (default: 0).

validation_frame¶

Id of the validation data frame.

Type: str.

weights_column¶

Column with observation weights. Giving some observation a weight of zero is equivalent to excluding it from the dataset; giving an observation a relative weight of 2 is equivalent to repeating that row twice. Negative weights are not allowed.

Type: str.

Bases: h2o.estimators.estimator_base.H2OEstimator

Naive Bayes

The naive Bayes classifier assumes independence between predictor variables conditional on the response, and a Gaussian distribution of numeric predictors with mean and standard deviation computed from the training dataset. When building a naive Bayes classifier, every row in the training dataset that contains at least one NA will be skipped completely. If the test dataset has missing values, then those predictors are omitted in the probability calculation during prediction.

balance_classes¶

Balance training data class counts via over/under-sampling (for imbalanced data).

Type: bool (default: False).

class_sampling_factors¶

Desired over/under-sampling ratios per class (in lexicographic order). If not specified, sampling factors will be automatically computed to obtain class balance during training. Requires balance_classes.

Type: List[float].

compute_metrics¶

Compute metrics on training data

Type: bool (default: True).

eps_prob¶

Cutoff below which probability is replaced with min_prob

Type: float (default: 0).

eps_sdev¶

Cutoff below which standard deviation is replaced with min_sdev

Type: float (default: 0).

fold_assignment¶

Cross-validation fold assignment scheme, if fold_column is not specified. The ‘Stratified’ option will stratify the folds based on the response variable, for classification problems.

One of: "auto", "random", "modulo", "stratified" (default: "auto").

fold_column¶

Column with cross-validation fold index assignment per observation.

Type: str.

ignore_const_cols¶

Ignore constant columns.

Type: bool (default: True).

ignored_columns¶

Names of columns to ignore for training.

Type: List[str].

keep_cross_validation_fold_assignment¶

Whether to keep the cross-validation fold assignment.

Type: bool (default: False).

keep_cross_validation_predictions¶

Whether to keep the predictions of the cross-validation models.

Type: bool (default: False).

laplace¶

Laplace smoothing parameter

Type: float (default: 0).

max_after_balance_size¶

Maximum relative size of the training data after balancing class counts (can be less than 1.0). Requires balance_classes.

Type: float (default: 5).

max_confusion_matrix_size¶

[Deprecated] Maximum size (# classes) for confusion matrices to be printed in the Logs

Type: int (default: 20).

max_hit_ratio_k¶

Max. number (top K) of predictions to use for hit ratio computation (for multi-class only, 0 to disable)

Type: int (default: 0).

max_runtime_secs¶

Maximum allowed runtime in seconds for model training. Use 0 to disable.

Type: float (default: 0).

min_prob¶

Min. probability to use for observations with not enough data

Type: float (default: 0.001).

min_sdev¶

Min. standard deviation to use for observations with not enough data

Type: float (default: 0.001).

nfolds¶

Number of folds for N-fold cross-validation (0 to disable or >= 2).

Type: int (default: 0).

response_column¶

Response variable column.

Type: str.

score_each_iteration¶

Whether to score during each iteration of model training.

Type: bool (default: False).

seed¶

Seed for pseudo random number generator (only used for cross-validation and fold_assignment=”Random” or “AUTO”)

Type: int (default: -1).

training_frame¶

Id of the training data frame (Not required, to allow initial validation of model parameters).

Type: str.

validation_frame¶

Id of the validation data frame.

Type: str.

Bases: h2o.estimators.estimator_base.H2OEstimator

Distributed Random Forest

balance_classes¶

Balance training data class counts via over/under-sampling (for imbalanced data).

Type: bool (default: False).

binomial_double_trees¶

For binary classification: Build 2x as many trees (one per class) - can lead to higher accuracy.

Type: bool (default: False).

build_tree_one_node¶

Run on one node only; no network overhead but fewer cpus used. Suitable for small datasets.

Type: bool (default: False).

categorical_encoding¶

Encoding scheme for categorical features

One of: "auto", "enum", "one_hot_internal", "one_hot_explicit", "binary", "eigen" (default: "auto").

checkpoint¶

Model checkpoint to resume training with.

Type: str.

class_sampling_factors¶

Desired over/under-sampling ratios per class (in lexicographic order). If not specified, sampling factors will be automatically computed to obtain class balance during training. Requires balance_classes.

Type: List[float].

col_sample_rate_change_per_level¶

Relative change of the column sampling rate for every level (from 0.0 to 2.0)

Type: float (default: 1).

col_sample_rate_per_tree¶

Column sample rate per tree (from 0.0 to 1.0)

Type: float (default: 1).

fold_assignment¶

Cross-validation fold assignment scheme, if fold_column is not specified. The ‘Stratified’ option will stratify the folds based on the response variable, for classification problems.

One of: "auto", "random", "modulo", "stratified" (default: "auto").

fold_column¶

Column with cross-validation fold index assignment per observation.

Type: str.

histogram_type¶

What type of histogram to use for finding optimal split points

One of: "auto", "uniform_adaptive", "random", "quantiles_global", "round_robin" (default: "auto").

ignore_const_cols¶

Ignore constant columns.

Type: bool (default: True).

ignored_columns¶

Names of columns to ignore for training.

Type: List[str].

keep_cross_validation_fold_assignment¶

Whether to keep the cross-validation fold assignment.

Type: bool (default: False).

keep_cross_validation_predictions¶

Whether to keep the predictions of the cross-validation models.

Type: bool (default: False).

max_after_balance_size¶

Maximum relative size of the training data after balancing class counts (can be less than 1.0). Requires balance_classes.

Type: float (default: 5).

max_confusion_matrix_size¶

[Deprecated] Maximum size (# classes) for confusion matrices to be printed in the Logs

Type: int (default: 20).

max_depth¶

Maximum tree depth.

Type: int (default: 20).

max_hit_ratio_k¶

Max. number (top K) of predictions to use for hit ratio computation (for multi-class only, 0 to disable)

Type: int (default: 0).

max_runtime_secs¶

Maximum allowed runtime in seconds for model training. Use 0 to disable.

Type: float (default: 0).

min_rows¶

Fewest allowed (weighted) observations in a leaf.

Type: float (default: 1).

min_split_improvement¶

Minimum relative improvement in squared error reduction for a split to happen

Type: float (default: 1e-05).

mtries¶

Number of variables randomly sampled as candidates at each split. If set to -1, defaults to sqrt{p} for classification and p/3 for regression (where p is the # of predictors

Type: int (default: -1).

nbins¶

For numerical columns (real/int), build a histogram of (at least) this many bins, then split at the best point

Type: int (default: 20).

nbins_cats¶

For categorical columns (factors), build a histogram of this many bins, then split at the best point. Higher values can lead to more overfitting.

Type: int (default: 1024).

nbins_top_level¶

For numerical columns (real/int), build a histogram of (at most) this many bins at the root level, then decrease by factor of two per level

Type: int (default: 1024).

nfolds¶

Number of folds for N-fold cross-validation (0 to disable or >= 2).

Type: int (default: 0).

ntrees¶

Number of trees.

Type: int (default: 50).

offset_column¶

Offset column. This will be added to the combination of columns before applying the link function.

Type: str.

r2_stopping¶

r2_stopping is no longer supported and will be ignored if set - please use stopping_rounds, stopping_metric and stopping_tolerance instead. Previous version of H2O would stop making trees when the R^2 metric equals or exceeds this

Type: float (default: 1.797693135e+308).

response_column¶

Response variable column.

Type: str.

sample_rate¶

Row sample rate per tree (from 0.0 to 1.0)

Type: float (default: 0.6320000291).

sample_rate_per_class¶

Row sample rate per tree per class (from 0.0 to 1.0)

Type: List[float].

score_each_iteration¶

Whether to score during each iteration of model training.

Type: bool (default: False).

score_tree_interval¶

Score the model after every so many trees. Disabled if set to 0.

Type: int (default: 0).

seed¶

Seed for pseudo random number generator (if applicable)

Type: int (default: -1).

stopping_metric¶

Metric to use for early stopping (AUTO: logloss for classification, deviance for regression)

One of: "auto", "deviance", "logloss", "mse", "rmse", "mae", "rmsle", "auc", "lift_top_group", "misclassification", "mean_per_class_error" (default: "auto").

stopping_rounds¶

Early stopping based on convergence of stopping_metric. Stop if simple moving average of length k of the stopping_metric does not improve for k:=stopping_rounds scoring events (0 to disable)

Type: int (default: 0).

stopping_tolerance¶

Relative tolerance for metric-based stopping criterion (stop if relative improvement is not at least this much)

Type: float (default: 0.001).

training_frame¶

Id of the training data frame (Not required, to allow initial validation of model parameters).

Type: str.

validation_frame¶

Id of the validation data frame.

Type: str.

weights_column¶

Column with observation weights. Giving some observation a weight of zero is equivalent to excluding it from the dataset; giving an observation a relative weight of 2 is equivalent to repeating that row twice. Negative weights are not allowed.

Type: str.

Bases: h2o.estimators.estimator_base.H2OEstimator

Aggregator

categorical_encoding¶

Encoding scheme for categorical features

One of: "auto", "enum", "one_hot_internal", "one_hot_explicit", "binary", "eigen" (default: "auto").

ignore_const_cols¶

Ignore constant columns.

Type: bool (default: True).

ignored_columns¶

Names of columns to ignore for training.

Type: List[str].

radius_scale¶

Radius scaling

Type: float (default: 1).

response_column¶

Response variable column.

Type: str.

training_frame¶

Id of the training data frame (Not required, to allow initial validation of model parameters).

Type: str.

transform¶

Transformation of training data

One of: "none", "standardize", "normalize", "demean", "descale" (default: "normalize").

Bases: h2o.estimators.deeplearning.H2ODeepLearningEstimator

Examples

>>> import h2o as ml
>>> from h2o.estimators.deeplearning import H2OAutoEncoderEstimator
>>> ml.init()
>>> rows = [[1,2,3,4,0]*50, [2,1,2,4,1]*50, [2,1,4,2,1]*50, [0,1,2,34,1]*50, [2,3,4,1,0]*50]
>>> fr = ml.H2OFrame(rows)
>>> fr[4] = fr[4].asfactor()
>>> model = H2OAutoEncoderEstimator()
>>> model.train(x=range(4), training_frame=fr)

Bases: h2o.estimators.estimator_base.H2OEstimator

Generalized Low Rank Modeling

Builds a generalized low rank model of a H2O dataset.

expand_user_y¶

Expand categorical columns in user-specified initial Y

Type: bool (default: True).

gamma_x¶

Regularization weight on X matrix

Type: float (default: 0).

gamma_y¶

Regularization weight on Y matrix

Type: float (default: 0).

ignore_const_cols¶

Ignore constant columns.

Type: bool (default: True).

ignored_columns¶

Names of columns to ignore for training.

Type: List[str].

impute_original¶

Reconstruct original training data by reversing transform

Type: bool (default: False).

init¶

Initialization mode

One of: "random", "svd", "plus_plus", "user" (default: "plus_plus").

init_step_size¶

Initial step size

Type: float (default: 1).

k¶

Rank of matrix approximation

Type: int (default: 1).

loading_name¶

Frame key to save resulting X

Type: str.

loss¶

Numeric loss function

One of: "quadratic", "absolute", "huber", "poisson", "hinge", "logistic", "periodic" (default: "quadratic").

loss_by_col¶

Loss function by column (override)

Type: List[Enum["quadratic", "absolute", "huber", "poisson", "hinge", "logistic", "periodic", "categorical", "ordinal"]].

loss_by_col_idx¶

Loss function by column index (override)

Type: List[int].

max_iterations¶

Maximum number of iterations

Type: int (default: 1000).

max_runtime_secs¶

Maximum allowed runtime in seconds for model training. Use 0 to disable.

Type: float (default: 0).

max_updates¶

Maximum number of updates, defaults to 2*max_iterations

Type: int (default: 2000).

min_step_size¶

Minimum step size

Type: float (default: 0.0001).

multi_loss¶

Categorical loss function

One of: "categorical", "ordinal" (default: "categorical").

period¶

Length of period (only used with periodic loss function)

Type: int (default: 1).

recover_svd¶

Recover singular values and eigenvectors of XY

Type: bool (default: False).

regularization_x¶

Regularization function for X matrix

One of: "none", "quadratic", "l2", "l1", "non_negative", "one_sparse", "unit_one_sparse", "simplex" (default: "none").

regularization_y¶

Regularization function for Y matrix

One of: "none", "quadratic", "l2", "l1", "non_negative", "one_sparse", "unit_one_sparse", "simplex" (default: "none").

score_each_iteration¶

Whether to score during each iteration of model training.

Type: bool (default: False).

seed¶

RNG seed for initialization

Type: int (default: -1).

svd_method¶

Method for computing SVD during initialization (Caution: Power and Randomized are currently experimental and unstable)

One of: "gram_s_v_d", "power", "randomized" (default: "randomized").

training_frame¶

Id of the training data frame (Not required, to allow initial validation of model parameters).

Type: str.

transform¶

Transformation of training data

One of: "none", "standardize", "normalize", "demean", "descale" (default: "none").

user_x¶

User-specified initial X

Type: str.

user_y¶

User-specified initial Y

Type: str.

validation_frame¶

Id of the validation data frame.

Type: str.

Bases: h2o.estimators.estimator_base.H2OEstimator

K-means

Performs k-means clustering on an H2O dataset.

categorical_encoding¶

Encoding scheme for categorical features

One of: "auto", "enum", "one_hot_internal", "one_hot_explicit", "binary", "eigen" (default: "auto").

estimate_k¶

Whether to estimate the number of clusters (<=k) iteratively and deterministically.

Type: bool (default: False).

fold_assignment¶

Cross-validation fold assignment scheme, if fold_column is not specified. The ‘Stratified’ option will stratify the folds based on the response variable, for classification problems.

One of: "auto", "random", "modulo", "stratified" (default: "auto").

fold_column¶

Column with cross-validation fold index assignment per observation.

Type: str.

ignore_const_cols¶

Ignore constant columns.

Type: bool (default: True).

ignored_columns¶

Names of columns to ignore for training.

Type: List[str].

init¶

Initialization mode

One of: "random", "plus_plus", "furthest", "user" (default: "furthest").

k¶

The max. number of clusters. If estimate_k is disabled, the model will find k centroids, otherwise it will find up to k centroids.

Type: int (default: 1).

keep_cross_validation_fold_assignment¶

Whether to keep the cross-validation fold assignment.

Type: bool (default: False).

keep_cross_validation_predictions¶

Whether to keep the predictions of the cross-validation models.

Type: bool (default: False).

max_iterations¶

Maximum training iterations (if estimate_k is enabled, then this is for each inner Lloyds iteration)

Type: int (default: 10).

max_runtime_secs¶

Maximum allowed runtime in seconds for model training. Use 0 to disable.

Type: float (default: 0).

nfolds¶

Number of folds for N-fold cross-validation (0 to disable or >= 2).

Type: int (default: 0).

score_each_iteration¶

Whether to score during each iteration of model training.

Type: bool (default: False).

seed¶

RNG Seed

Type: int (default: -1).

standardize¶

Standardize columns before computing distances

Type: bool (default: True).

training_frame¶

Id of the training data frame (Not required, to allow initial validation of model parameters).

Type: str.

user_points¶

User-specified points

Type: str.

validation_frame¶

Id of the validation data frame.

Type: str.

Bases: h2o.estimators.estimator_base.H2OEstimator

Principal Components Analysis

compute_metrics¶

Whether to compute metrics on the training data

Type: bool (default: True).

ignore_const_cols¶

Ignore constant columns.

Type: bool (default: True).

ignored_columns¶

Names of columns to ignore for training.

Type: List[str].

impute_missing¶

Whether to impute missing entries with the column mean

Type: bool (default: False).

k¶

Rank of matrix approximation

Type: int (default: 1).

max_iterations¶

Maximum training iterations

Type: int (default: 1000).

max_runtime_secs¶

Maximum allowed runtime in seconds for model training. Use 0 to disable.

Type: float (default: 0).

pca_method¶

Method for computing PCA (Caution: Power and GLRM are currently experimental and unstable)

One of: "gram_s_v_d", "power", "randomized", "glrm" (default: "gram_s_v_d").

score_each_iteration¶

Whether to score during each iteration of model training.

Type: bool (default: False).

seed¶

RNG seed for initialization

Type: int (default: -1).

training_frame¶

Id of the training data frame (Not required, to allow initial validation of model parameters).

Type: str.

transform¶

Transformation of training data

One of: "none", "standardize", "normalize", "demean", "descale" (default: "none").

use_all_factor_levels¶

Whether first factor level is included in each categorical expansion

Type: bool (default: False).

validation_frame¶

Id of the validation data frame.

Type: str.

Bases: h2o.estimators.estimator_base.H2OEstimator

Stacked Ensemble

Builds a stacked ensemble (aka “super learner”) machine learning method that uses two or more H2O learning algorithms to improve predictive performance. It is a loss-based supervised learning method that finds the optimal combination of a collection of prediction algorithms.This method supports regression and binary classification.

Examples

>>> import h2o
>>> h2o.init()
>>> from h2o.estimators.random_forest import H2ORandomForestEstimator
>>> from h2o.estimators.gbm import H2OGradientBoostingEstimator
>>> from h2o.estimators.stackedensemble import H2OStackedEnsembleEstimator
>>> col_types = ["numeric", "numeric", "numeric", "enum", "enum", "numeric", "numeric", "numeric", "numeric"]
>>> dat = h2o.import_file("http://h2o-public-test-data.s3.amazonaws.com/smalldata/prostate/prostate.csv", destination_frame="prostate_hex", col_types= col_types)
>>> train, test = dat.split_frame(ratios=[.8], seed=1)
>>> x = ["CAPSULE","GLEASON","RACE","DPROS","DCAPS","PSA","VOL"]
>>> y = "AGE"
>>> my_gbm = H2OGradientBoostingEstimator()
>>> my_gbm.train(x=x, y=y, training_frame=train)
>>> my_rf = H2ORandomForestEstimator()
>>> my_rf.train(x=x, y=y, training_frame=train)
>>> stack = H2OStackedEnsembleEstimator(model_id="my_ensemble_guassian", training_frame=train, validation_frame=test, base_models=[my_gbm.model_id,  my_rf.model_id], selection_strategy="choose_all")
>>> stack.train(x=x, y=y, training_frame=train, validation_frame=test)
>>> stack.model_performance()

base_models¶

List of model ids which we can stack together. Which ones are chosen depends on the selection_strategy (currently, all models will be used since selection_strategy can only be set to choose_all). Models must have been cross-validated using nfolds > 1, fold_assignment equal to Modulo, and keep_cross_validation_folds must be set to True.

Type: List[str] (default: []).

selection_strategy¶

Strategy for choosing which models to stack.

One of: "choose_all".

training_frame¶

Id of the training data frame (Not required, to allow initial validation of model parameters).

Type: str.

validation_frame¶

Id of the validation data frame.

Type: str.

Bases: h2o.model.model_base.ModelBase

Base class for H2O Estimators.

H2O Estimators implement the following methods for model construction:

• start() - Top-level user-facing API for asynchronous model build
• join() - Top-level user-facing API for blocking on async model build
• train() - Top-level user-facing API for model building.
• fit() - Used by scikit-learn.

Because H2OEstimator instances are instances of ModelBase, these objects can use the H2O model API.

fit(x, y=None, **params)[source]¶

Fit an H2O model as part of a scikit-learn pipeline or grid search.

A warning will be issued if a caller other than sklearn attempts to use this method.

Parameters:	x (H2OFrame) – An H2OFrame consisting of the predictor variables. y (H2OFrame) – An H2OFrame consisting of the response variable. params – Extra arguments.
Returns:	The current instance of H2OEstimator for method chaining.

get_params(deep=True)[source]¶

Obtain parameters for this estimator.

Used primarily for sklearn Pipelines and sklearn grid search.

Parameters:	deep – If True, return parameters of all sub-objects that are estimators.
Returns:	A dict of parameters

join()[source]¶

Wait until job’s completion.

set_params(**parms)[source]¶

Used by sklearn for updating parameters during grid search.

Parameters:	parms – A dictionary of parameters that will be set on this model.
Returns:	self, the current estimator object with the parameters all set as desired.

start(x, y=None, training_frame=None, offset_column=None, fold_column=None, weights_column=None, validation_frame=None, **params)[source]¶

Train the model asynchronously (to block for results call join()).

Parameters:

x – A list of column names or indices indicating the predictor columns. y – An index or a column name indicating the response column. training_frame (H2OFrame) – The H2OFrame having the columns indicated by x and y (as well as any additional columns specified by fold, offset, and weights). offset_column – The name or index of the column in training_frame that holds the offsets. fold_column – The name or index of the column in training_frame that holds the per-row fold assignments. weights_column – The name or index of the column in training_frame that holds the per-row weights. validation_frame – H2OFrame with validation data to be scored on while training.

train(x=None, y=None, training_frame=None, offset_column=None, fold_column=None, weights_column=None, validation_frame=None, max_runtime_secs=None, ignored_columns=None, model_id=None)[source]¶

Train the H2O model.

Parameters:

x – A list of column names or indices indicating the predictor columns. y – An index or a column name indicating the response column. training_frame (H2OFrame) – The H2OFrame having the columns indicated by x and y (as well as any additional columns specified by fold, offset, and weights). offset_column – The name or index of the column in training_frame that holds the offsets. fold_column – The name or index of the column in training_frame that holds the per-row fold assignments. weights_column – The name or index of the column in training_frame that holds the per-row weights. validation_frame – H2OFrame with validation data to be scored on while training. max_runtime_secs (float) – Maximum allowed runtime in seconds for model training. Use 0 to disable.

Bases: h2o.utils.backward_compatibility.BackwardsCompatibleBase

Grid Search of a Hyper-Parameter Space for a Model

Examples

>>> from h2o.grid.grid_search import H2OGridSearch
>>> from h2o.estimators.glm import H2OGeneralizedLinearEstimator
>>> hyper_parameters = {'alpha': [0.01,0.5], 'lambda': [1e-5,1e-6]}
>>> gs = H2OGridSearch(H2OGeneralizedLinearEstimator(family='binomial'), hyper_parameters)
>>> training_data = h2o.import_file("smalldata/logreg/benign.csv")
>>> gs.train(x=range(3) + range(4,11),y=3, training_frame=training_data)
>>> gs.show()

aic(train=False, valid=False, xval=False)[source]¶

Get the AIC(s).

If all are False (default), then return the training metric value. If more than one options is set to True, then return a dictionary of metrics where the keys are “train”, “valid”, and “xval”.

Parameters:	train (bool) – If train is True, then return the AIC value for the training data. valid (bool) – If valid is True, then return the AIC value for the validation data. xval (bool) – If xval is True, then return the AIC value for the validation data.
Returns:	The AIC.

auc(train=False, valid=False, xval=False)[source]¶

Get the AUC(s).

If all are False (default), then return the training metric value. If more than one options is set to True, then return a dictionary of metrics where the keys are “train”, “valid”, and “xval”.

Parameters:	train (bool) – If train is True, then return the AUC value for the training data. valid (bool) – If valid is True, then return the AUC value for the validation data. xval (bool) – If xval is True, then return the AUC value for the validation data.
Returns:	The AUC.

biases(vector_id=0)[source]¶

Return the frame for the respective bias vector.

Param:	vector_id: an integer, ranging from 0 to number of layers, that specifies the bias vector to return.
Returns:	an H2OFrame which represents the bias vector identified by vector_id

build_model(algo_params)[source]¶

(internal)

catoffsets()[source]¶

Categorical offsets for one-hot encoding

coef()[source]¶

Return the coefficients for this model.

coef_norm()[source]¶

Return the normalized coefficients.

deepfeatures(test_data, layer)[source]¶

Obtain a hidden layer’s details on a dataset.

Parameters:	test_data – Data to create a feature space on. layer (int) – Index of the hidden layer.
Returns:	A dictionary of hidden layer details for each model.

get_grid(sort_by=None, decreasing=None)[source]¶

Retrieve an H2OGridSearch instance.

Optionally specify a metric by which to sort models and a sort order. Note that if neither cross-validation nor a validation frame is used in the grid search, then the training metrics will display in the “get grid” output. If a validation frame is passed to the grid, and nfolds = 0, then the validation metrics will display. However, if nfolds > 1, then cross-validation metrics will display even if a validation frame is provided.

Parameters:	sort_by (str) – A metric by which to sort the models in the grid space. Choices are: "logloss", "residual_deviance", "mse", "auc", "r2", "accuracy", "precision", "recall", "f1", etc. decreasing (bool) – Sort the models in decreasing order of metric if true, otherwise sort in increasing order (default).
Returns:	A new H2OGridSearch instance optionally sorted on the specified metric.

get_hyperparams(id, display=True)[source]¶

Get the hyperparameters of a model explored by grid search.

Parameters:	id (str) – The model id of the model with hyperparameters of interest. display (bool) – Flag to indicate whether to display the hyperparameter names.
Returns:	A list of the hyperparameters for the specified model.

get_hyperparams_dict(id, display=True)[source]¶

Derived and returned the model parameters used to train the particular grid search model.

Parameters:	id (str) – The model id of the model with hyperparameters of interest. display (bool) – Flag to indicate whether to display the hyperparameter names.
Returns:	A dict of model pararmeters derived from the hyper-parameters used to train this particular model.

get_xval_models(key=None)[source]¶

Return a Model object.

Parameters:	key (str) – If None, return all cross-validated models; otherwise return the model specified by the key.
Returns:	A model or a list of models.

gini(train=False, valid=False, xval=False)[source]¶

Get the Gini Coefficient(s).

If all are False (default), then return the training metric value. If more than one options is set to True, then return a dictionary of metrics where the keys are “train”, “valid”, and “xval”.

Parameters:	train (bool) – If train is True, then return the Gini Coefficient value for the training data. valid (bool) – If valid is True, then return the Gini Coefficient value for the validation data. xval (bool) – If xval is True, then return the Gini Coefficient value for the cross validation data.
Returns:	The Gini Coefficient for this binomial model.

grid_id¶

A key that identifies this grid search object in H2O.

is_cross_validated()[source]¶

Return True if the model was cross-validated.

join()[source]¶

Wait until grid finishes computing.

logloss(train=False, valid=False, xval=False)[source]¶

Get the Log Loss(s).

If all are False (default), then return the training metric value. If more than one options is set to True, then return a dictionary of metrics where the keys are “train”, “valid”, and “xval”.

Parameters:	train (bool) – If train is True, then return the Log Loss value for the training data. valid (bool) – If valid is True, then return the Log Loss value for the validation data. xval (bool) – If xval is True, then return the Log Loss value for the cross validation data.
Returns:	The Log Loss for this binomial model.

mean_residual_deviance(train=False, valid=False, xval=False)[source]¶

Get the Mean Residual Deviances(s).

If all are False (default), then return the training metric value. If more than one options is set to True, then return a dictionary of metrics where the keys are “train”, “valid”, and “xval”.

Parameters:	train (bool) – If train is True, then return the Mean Residual Deviance value for the training data. valid (bool) – If valid is True, then return the Mean Residual Deviance value for the validation data. xval (bool) – If xval is True, then return the Mean Residual Deviance value for the cross validation data.
Returns:	The Mean Residual Deviance for this regression model.

model_performance(test_data=None, train=False, valid=False, xval=False)[source]¶

Generate model metrics for this model on test_data.

Parameters:	test_data – Data set for which model metrics shall be computed against. All three of train, valid and xval arguments are ignored if test_data is not None. train – Report the training metrics for the model. valid – Report the validation metrics for the model. xval – Report the validation metrics for the model.
Returns:	An object of class H2OModelMetrics.

mse(train=False, valid=False, xval=False)[source]¶

Get the MSE(s).

If all are False (default), then return the training metric value. If more than one options is set to True, then return a dictionary of metrics where the keys are “train”, “valid”, and “xval”.

Parameters:	train (bool) – If train is True, then return the MSE value for the training data. valid (bool) – If valid is True, then return the MSE value for the validation data. xval (bool) – If xval is True, then return the MSE value for the cross validation data.
Returns:	The MSE for this regression model.

normmul()[source]¶

Normalization/Standardization multipliers for numeric predictors.

normsub()[source]¶

Normalization/Standardization offsets for numeric predictors.

null_degrees_of_freedom(train=False, valid=False, xval=False)[source]¶

Retreive the null degress of freedom if this model has the attribute, or None otherwise.

Parameters:	train (bool) – Get the null dof for the training set. If both train and valid are False, then train is selected by default. valid (bool) – Get the null dof for the validation set. If both train and valid are True, then train is selected by default. xval (bool) – Get the null dof for the cross-validated models.
Returns:	the null dof, or None if it is not present.

null_deviance(train=False, valid=False, xval=False)[source]¶

Retreive the null deviance if this model has the attribute, or None otherwise.

Parameters:	train (bool) – Get the null deviance for the training set. If both train and valid are False, then train is selected by default. valid (bool) – Get the null deviance for the validation set. If both train and valid are True, then train is selected by default. xval (bool) – Get the null deviance for the cross-validated models.
Returns:	the null deviance, or None if it is not present.

pprint_coef()[source]¶

Pretty print the coefficents table (includes normalized coefficients).

predict(test_data)[source]¶

Predict on a dataset.

Parameters:	test_data (H2OFrame) – Data to be predicted on.
Returns:	H2OFrame filled with predictions.

r2(train=False, valid=False, xval=False)[source]¶

Return the R^2 for this regression model.

The R^2 value is defined to be 1 - MSE/var, where var is computed as sigma^2.

If all are False (default), then return the training metric value. If more than one options is set to True, then return a dictionary of metrics where the keys are “train”, “valid”, and “xval”.

Parameters:	train (bool) – If train is True, then return the R^2 value for the training data. valid (bool) – If valid is True, then return the R^2 value for the validation data. xval (bool) – If xval is True, then return the R^2 value for the cross validation data.
Returns:	The R^2 for this regression model.

residual_degrees_of_freedom(train=False, valid=False, xval=False)[source]¶

Retreive the residual degress of freedom if this model has the attribute, or None otherwise.

Parameters:	train (bool) – Get the residual dof for the training set. If both train and valid are False, then train is selected by default. valid (bool) – Get the residual dof for the validation set. If both train and valid are True, then train is selected by default. xval (bool) – Get the residual dof for the cross-validated models.
Returns:	the residual degrees of freedom, or None if they are not present.

residual_deviance(train=False, valid=False, xval=False)[source]¶

Retreive the residual deviance if this model has the attribute, or None otherwise.

Parameters:	train (bool) – Get the residual deviance for the training set. If both train and valid are False, then train is selected by default. valid (bool) – Get the residual deviance for the validation set. If both train and valid are True, then train is selected by default. xval (bool) – Get the residual deviance for the cross-validated models.
Returns:	the residual deviance, or None if it is not present.

respmul()[source]¶

Normalization/Standardization multipliers for numeric response.

respsub()[source]¶

Normalization/Standardization offsets for numeric response.

scoring_history()[source]¶

Retrieve model scoring history.

Returns:	Score history (H2OTwoDimTable)

show()[source]¶

Print models sorted by metric.

sort_by(*args, **kwargs)[source]¶

grid.sort_by() is deprecated; use grid.get_grid() instead

sorted_metric_table()[source]¶

Retrieve summary table of an H2O Grid Search.

Returns:	The summary table as an H2OTwoDimTable or a Pandas DataFrame.

start(x, y=None, training_frame=None, offset_column=None, fold_column=None, weights_column=None, validation_frame=None, **params)[source]¶

Asynchronous model build by specifying the predictor columns, response column, and any additional frame-specific values.

To block for results, call join().

Parameters:

x – A list of column names or indices indicating the predictor columns. y – An index or a column name indicating the response column. training_frame – The H2OFrame having the columns indicated by x and y (as well as any additional columns specified by fold, offset, and weights). offset_column – The name or index of the column in training_frame that holds the offsets. fold_column – The name or index of the column in training_frame that holds the per-row fold assignments. weights_column – The name or index of the column in training_frame that holds the per-row weights. validation_frame – H2OFrame with validation data to be scored on while training.

summary(header=True)[source]¶

Print a detailed summary of the explored models.

train(x, y=None, training_frame=None, offset_column=None, fold_column=None, weights_column=None, validation_frame=None, **params)[source]¶

Train the model synchronously (i.e. do not return until the model finishes training).

To train asynchronously call start().

Parameters:

x – A list of column names or indices indicating the predictor columns. y – An index or a column name indicating the response column. training_frame – The H2OFrame having the columns indicated by x and y (as well as any additional columns specified by fold, offset, and weights). offset_column – The name or index of the column in training_frame that holds the offsets. fold_column – The name or index of the column in training_frame that holds the per-row fold assignments. weights_column – The name or index of the column in training_frame that holds the per-row weights. validation_frame – H2OFrame with validation data to be scored on while training.

varimp(use_pandas=False)[source]¶

Pretty print the variable importances, or return them in a list/pandas DataFrame.

Parameters:	use_pandas (bool) – If True, then the variable importances will be returned as a pandas data frame.
Returns:	A dictionary of lists or Pandas DataFrame instances.

weights(matrix_id=0)[source]¶

Return the frame for the respective weight matrix.

Param:	matrix_id: an integer, ranging from 0 to number of layers, that specifies the weight matrix to return.
Returns:	an H2OFrame which represents the weight matrix identified by matrix_id

xval_keys()[source]¶

Model keys for the cross-validated model.

xvals()[source]¶

Return the list of cross-validated models.

Bases: h2o.estimators.estimator_base.H2OEstimator

Singular Value Decomposition

ignore_const_cols¶

Ignore constant columns.

Type: bool (default: True).

ignored_columns¶

Names of columns to ignore for training.

Type: List[str].

keep_u¶

Save left singular vectors?

Type: bool (default: True).

max_iterations¶

Maximum iterations

Type: int (default: 1000).

max_runtime_secs¶

Maximum allowed runtime in seconds for model training. Use 0 to disable.

Type: float (default: 0).

nv¶

Number of right singular vectors

Type: int (default: 1).

score_each_iteration¶

Whether to score during each iteration of model training.

Type: bool (default: False).

seed¶

RNG seed for k-means++ initialization

Type: int (default: -1).

svd_method¶

Method for computing SVD (Caution: Power and Randomized are currently experimental and unstable)

One of: "gram_s_v_d", "power", "randomized" (default: "gram_s_v_d").

training_frame¶

Id of the training data frame (Not required, to allow initial validation of model parameters).

Type: str.

transform¶

Transformation of training data

One of: "none", "standardize", "normalize", "demean", "descale" (default: "none").

u_name¶

Frame key to save left singular vectors

Type: str.

use_all_factor_levels¶

Whether first factor level is included in each categorical expansion

Type: bool (default: True).

validation_frame¶

Id of the validation data frame.

Type: str.

Bases: h2o.estimators.estimator_base.H2OEstimator

Word2Vec

epochs¶

Number of training iterations to run

Type: int (default: 5).

init_learning_rate¶

Set the starting learning rate

Type: float (default: 0.025).

min_word_freq¶

This will discard words that appear less than <int> times

Type: int (default: 5).

norm_model¶

Use Hierarchical Softmax

One of: "hsm" (default: "hsm").

sent_sample_rate¶

Set threshold for occurrence of words. Those that appear with higher frequency in the training data will be randomly down-sampled; useful range is (0, 1e-5)

Type: float (default: 0.001).

training_frame¶

Id of the training data frame (Not required, to allow initial validation of model parameters).

Type: str.

vec_size¶

Set size of word vectors

Type: int (default: 100).

window_size¶

Set max skip length between words

Type: int (default: 5).

word_model¶

Use the Skip-Gram model

One of: "skip_gram" (default: "skip_gram").