DeepLearningParametersV3 (h2o-java-rest-bindings version 3.6.0.12 API)

java.lang.Object
- water.bindings.pojos.Schema
- - water.bindings.pojos.ModelParametersSchema
  - - water.bindings.pojos.DeepLearningParametersV3

public class DeepLearningParametersV3
extends ModelParametersSchema

Field Summary

Fields
Modifier and Type	Field and Description
`Activation`	`activation` Activation function
`boolean`	`adaptive_rate` Adaptive learning rate
`boolean`	`autoencoder` Auto-Encoder
`double`	`average_activation` Average activation for sparse auto-encoder (Experimental)
`boolean`	`balance_classes` Balance training data class counts via over/under-sampling (for imbalanced data).
`java.lang.String`	`checkpoint` Model checkpoint to resume training with
`float[]`	`class_sampling_factors` Desired over/under-sampling ratios per class (in lexicographic order).
`double`	`classification_stop` Stopping criterion for classification error fraction on training data (-1 to disable)
`boolean`	`col_major` Use a column major weight matrix for input layer.
`boolean`	`diagnostics` Enable diagnostics for hidden layers
`Family`	`distribution` Distribution function
`boolean`	`elastic_averaging` Elastic averaging between compute nodes can improve distributed model convergence (Experimental)
`double`	`elastic_averaging_moving_rate` Elastic averaging moving rate (only if elastic averaging is enabled).
`double`	`elastic_averaging_regularization` Elastic averaging regularization strength (only if elastic averaging is enabled).
`double`	`epochs` How many times the dataset should be iterated (streamed), can be fractional
`double`	`epsilon` Adaptive learning rate smoothing factor (to avoid divisions by zero and allow progress)
`boolean`	`export_weights_and_biases` Whether to export Neural Network weights and biases to H2O Frames
`boolean`	`fast_mode` Enable fast mode (minor approximation in back-propagation)
`FoldAssignmentScheme`	`fold_assignment` Cross-validation fold assignment scheme, if fold_column is not specified
`ColSpecifierV3`	`fold_column` Column with cross-validation fold index assignment per observation
`boolean`	`force_load_balance` Force extra load balancing to increase training speed for small datasets (to keep all cores busy)
`int[]`	`hidden` Hidden layer sizes (e.g.
`double[]`	`hidden_dropout_ratios` Hidden layer dropout ratios (can improve generalization), specify one value per hidden layer, defaults to 0.5
`boolean`	`ignore_const_cols` Ignore constant columns
`java.lang.String[]`	`ignored_columns` Ignored columns
`InitialWeightDistribution`	`initial_weight_distribution` Initial Weight Distribution
`double`	`initial_weight_scale` Uniform: -value...value, Normal: stddev)
`double`	`input_dropout_ratio` Input layer dropout ratio (can improve generalization, try 0.1 or 0.2)
`boolean`	`keep_cross_validation_predictions` Keep cross-validation model predictions
`double`	`l1` L1 regularization (can add stability and improve generalization, causes many weights to become 0)
`double`	`l2` L2 regularization (can add stability and improve generalization, causes many weights to be small
`Loss`	`loss` Loss function
`float`	`max_after_balance_size` Maximum relative size of the training data after balancing class counts (can be less than 1.0).
`int`	`max_categorical_features` Max.
`int`	`max_confusion_matrix_size` Maximum size (# classes) for confusion matrices to be printed in the Logs
`int`	`max_hit_ratio_k` Max.
`float`	`max_w2` Constraint for squared sum of incoming weights per unit (e.g.
`MissingValuesHandling`	`missing_values_handling` Handling of missing values.
`java.lang.String`	`model_id` Destination id for this model; auto-generated if not specified
`double`	`momentum_ramp` Number of training samples for which momentum increases
`double`	`momentum_stable` Final momentum after the ramp is over (try 0.99)
`double`	`momentum_start` Initial momentum at the beginning of training (try 0.5)
`boolean`	`nesterov_accelerated_gradient` Use Nesterov accelerated gradient (recommended)
`int`	`nfolds` Number of folds for N-fold cross-validation
`ColSpecifierV3`	`offset_column` Offset column
`boolean`	`overwrite_with_best_model` If enabled, override the final model with the best model found during training
`boolean`	`quiet_mode` Enable quiet mode for less output to standard output
`double`	`rate` Learning rate (higher => less stable, lower => slower convergence)
`double`	`rate_annealing` Learning rate annealing: rate / (1 + rate_annealing * samples)
`double`	`rate_decay` Learning rate decay factor between layers (N-th layer: rate*alpha^(N-1))
`double`	`regression_stop` Stopping criterion for regression error (MSE) on training data (-1 to disable)
`boolean`	`replicate_training_data` Replicate the entire training dataset onto every node for faster training on small datasets
`boolean`	`reproducible` Force reproducibility on small data (will be slow - only uses 1 thread)
`ColSpecifierV3`	`response_column` Response column
`double`	`rho` Adaptive learning rate time decay factor (similarity to prior updates)
`double`	`score_duty_cycle` Maximum duty cycle fraction for scoring (lower: more training, higher: more scoring).
`boolean`	`score_each_iteration` Whether to score during each iteration of model training
`double`	`score_interval` Shortest time interval (in secs) between model scoring
`long`	`score_training_samples` Number of training set samples for scoring (0 for all)
`long`	`score_validation_samples` Number of validation set samples for scoring (0 for all)
`ClassSamplingMethod`	`score_validation_sampling` Method used to sample validation dataset for scoring
`long`	`seed` Seed for random numbers (affects sampling) - Note: only reproducible when running single threaded
`boolean`	`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)
`boolean`	`single_node_mode` Run on a single node for fine-tuning of model parameters
`boolean`	`sparse` Sparse data handling (more efficient for data with lots of 0 values).
`double`	`sparsity_beta` Sparsity regularization (Experimental)
`StoppingMetric`	`stopping_metric` Metric to use for early stopping (AUTO: logloss for classification, deviance for regression)
`int`	`stopping_rounds` Early stopping based on convergence of stopping_metric.
`double`	`stopping_tolerance` Relative tolerance for metric-based stopping criterion Relative tolerance for metric-based stopping criterion (stop if relative improvement is not at least this much)
`double`	`target_ratio_comm_to_comp` Target ratio of communication overhead to computation.
`long`	`train_samples_per_iteration` Number of training samples (globally) per MapReduce iteration.
`java.lang.String`	`training_frame` Training frame
`double`	`tweedie_power` Tweedie Power
`boolean`	`use_all_factor_levels` Use all factor levels of categorical variables.
`java.lang.String`	`validation_frame` Validation frame
`boolean`	`variable_importances` Compute variable importances for input features (Gedeon method) - can be slow for large networks
`ColSpecifierV3`	`weights_column` Column with observation weights

Constructor Summary

Constructors
Constructor and Description

DeepLearningParametersV3()

Constructors
Constructor and Description
`DeepLearningParametersV3()`

Method Summary
- Methods inherited from class java.lang.Object
  clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

- Field Detail
  - balance_classes
```
public boolean balance_classes
```
    Balance training data class counts via over/under-sampling (for imbalanced data).
  - class_sampling_factors
```
public float[] 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.
  - max_after_balance_size
```
public float max_after_balance_size
```
    Maximum relative size of the training data after balancing class counts (can be less than 1.0). Requires balance_classes.
  - max_confusion_matrix_size
```
public int max_confusion_matrix_size
```
    Maximum size (# classes) for confusion matrices to be printed in the Logs
  - max_hit_ratio_k
```
public int max_hit_ratio_k
```
    Max. number (top K) of predictions to use for hit ratio computation (for multi-class only, 0 to disable)
  - overwrite_with_best_model
```
public boolean overwrite_with_best_model
```
    If enabled, override the final model with the best model found during training
  - autoencoder
```
public boolean autoencoder
```
    Auto-Encoder
  - use_all_factor_levels
```
public boolean 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.
  - activation
```
public Activation activation
```
    Activation function
  - hidden
```
public int[] hidden
```
    Hidden layer sizes (e.g. 100,100).
  - epochs
```
public double epochs
```
    How many times the dataset should be iterated (streamed), can be fractional
  - train_samples_per_iteration
```
public long 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
  - target_ratio_comm_to_comp
```
public double 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)
  - seed
```
public long seed
```
    Seed for random numbers (affects sampling) - Note: only reproducible when running single threaded
  - adaptive_rate
```
public boolean adaptive_rate
```
    Adaptive learning rate
  - rho
```
public double rho
```
    Adaptive learning rate time decay factor (similarity to prior updates)
  - epsilon
```
public double epsilon
```
    Adaptive learning rate smoothing factor (to avoid divisions by zero and allow progress)
  - rate
```
public double rate
```
    Learning rate (higher => less stable, lower => slower convergence)
  - rate_annealing
```
public double rate_annealing
```
    Learning rate annealing: rate / (1 + rate_annealing * samples)
  - rate_decay
```
public double rate_decay
```
    Learning rate decay factor between layers (N-th layer: rate*alpha^(N-1))
  - momentum_start
```
public double momentum_start
```
    Initial momentum at the beginning of training (try 0.5)
  - momentum_ramp
```
public double momentum_ramp
```
    Number of training samples for which momentum increases
  - momentum_stable
```
public double momentum_stable
```
    Final momentum after the ramp is over (try 0.99)
  - nesterov_accelerated_gradient
```
public boolean nesterov_accelerated_gradient
```
    Use Nesterov accelerated gradient (recommended)
  - input_dropout_ratio
```
public double input_dropout_ratio
```
    Input layer dropout ratio (can improve generalization, try 0.1 or 0.2)
  - hidden_dropout_ratios
```
public double[] hidden_dropout_ratios
```
    Hidden layer dropout ratios (can improve generalization), specify one value per hidden layer, defaults to 0.5
  - l1
```
public double l1
```
    L1 regularization (can add stability and improve generalization, causes many weights to become 0)
  - l2
```
public double l2
```
    L2 regularization (can add stability and improve generalization, causes many weights to be small
  - max_w2
```
public float max_w2
```
    Constraint for squared sum of incoming weights per unit (e.g. for Rectifier)
  - initial_weight_distribution
```
public InitialWeightDistribution initial_weight_distribution
```
    Initial Weight Distribution
  - initial_weight_scale
```
public double initial_weight_scale
```
    Uniform: -value...value, Normal: stddev)
  - loss
```
public Loss loss
```
    Loss function
  - distribution
```
public Family distribution
```
    Distribution function
  - tweedie_power
```
public double tweedie_power
```
    Tweedie Power
  - score_interval
```
public double score_interval
```
    Shortest time interval (in secs) between model scoring
  - score_training_samples
```
public long score_training_samples
```
    Number of training set samples for scoring (0 for all)
  - score_validation_samples
```
public long score_validation_samples
```
    Number of validation set samples for scoring (0 for all)
  - score_duty_cycle
```
public double score_duty_cycle
```
    Maximum duty cycle fraction for scoring (lower: more training, higher: more scoring).
  - classification_stop
```
public double classification_stop
```
    Stopping criterion for classification error fraction on training data (-1 to disable)
  - regression_stop
```
public double regression_stop
```
    Stopping criterion for regression error (MSE) on training data (-1 to disable)
  - quiet_mode
```
public boolean quiet_mode
```
    Enable quiet mode for less output to standard output
  - score_validation_sampling
```
public ClassSamplingMethod score_validation_sampling
```
    Method used to sample validation dataset for scoring
  - diagnostics
```
public boolean diagnostics
```
    Enable diagnostics for hidden layers
  - variable_importances
```
public boolean variable_importances
```
    Compute variable importances for input features (Gedeon method) - can be slow for large networks
  - fast_mode
```
public boolean fast_mode
```
    Enable fast mode (minor approximation in back-propagation)
  - force_load_balance
```
public boolean force_load_balance
```
    Force extra load balancing to increase training speed for small datasets (to keep all cores busy)
  - replicate_training_data
```
public boolean replicate_training_data
```
    Replicate the entire training dataset onto every node for faster training on small datasets
  - single_node_mode
```
public boolean single_node_mode
```
    Run on a single node for fine-tuning of model parameters
  - shuffle_training_data
```
public boolean 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)
  - missing_values_handling
```
public MissingValuesHandling missing_values_handling
```
    Handling of missing values. Either Skip or MeanImputation.
  - sparse
```
public boolean sparse
```
    Sparse data handling (more efficient for data with lots of 0 values).
  - col_major
```
public boolean col_major
```
    Use a column major weight matrix for input layer. Can speed up forward propagation, but might slow down backpropagation (Deprecated).
  - average_activation
```
public double average_activation
```
    Average activation for sparse auto-encoder (Experimental)
  - sparsity_beta
```
public double sparsity_beta
```
    Sparsity regularization (Experimental)
  - max_categorical_features
```
public int max_categorical_features
```
    Max. number of categorical features, enforced via hashing (Experimental)
  - reproducible
```
public boolean reproducible
```
    Force reproducibility on small data (will be slow - only uses 1 thread)
  - export_weights_and_biases
```
public boolean export_weights_and_biases
```
    Whether to export Neural Network weights and biases to H2O Frames
  - elastic_averaging
```
public boolean elastic_averaging
```
    Elastic averaging between compute nodes can improve distributed model convergence (Experimental)
  - elastic_averaging_moving_rate
```
public double elastic_averaging_moving_rate
```
    Elastic averaging moving rate (only if elastic averaging is enabled).
  - elastic_averaging_regularization
```
public double elastic_averaging_regularization
```
    Elastic averaging regularization strength (only if elastic averaging is enabled).
  - model_id
```
public java.lang.String model_id
```
    Destination id for this model; auto-generated if not specified
  - training_frame
```
public java.lang.String training_frame
```
    Training frame
  - validation_frame
```
public java.lang.String validation_frame
```
    Validation frame
  - nfolds
```
public int nfolds
```
    Number of folds for N-fold cross-validation
  - keep_cross_validation_predictions
```
public boolean keep_cross_validation_predictions
```
    Keep cross-validation model predictions
  - response_column
```
public ColSpecifierV3 response_column
```
    Response column
  - weights_column
```
public ColSpecifierV3 weights_column
```
    Column with observation weights
  - offset_column
```
public ColSpecifierV3 offset_column
```
    Offset column
  - fold_column
```
public ColSpecifierV3 fold_column
```
    Column with cross-validation fold index assignment per observation
  - fold_assignment
```
public FoldAssignmentScheme fold_assignment
```
    Cross-validation fold assignment scheme, if fold_column is not specified
  - ignored_columns
```
public java.lang.String[] ignored_columns
```
    Ignored columns
  - ignore_const_cols
```
public boolean ignore_const_cols
```
    Ignore constant columns
  - score_each_iteration
```
public boolean score_each_iteration
```
    Whether to score during each iteration of model training
  - checkpoint
```
public java.lang.String checkpoint
```
    Model checkpoint to resume training with
  - stopping_rounds
```
public int 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)
  - stopping_metric
```
public StoppingMetric stopping_metric
```
    Metric to use for early stopping (AUTO: logloss for classification, deviance for regression)
  - stopping_tolerance
```
public double stopping_tolerance
```
    Relative tolerance for metric-based stopping criterion Relative tolerance for metric-based stopping criterion (stop if relative improvement is not at least this much)
- Constructor Detail
  - DeepLearningParametersV3
```
public DeepLearningParametersV3()
```

Class DeepLearningParametersV3

Field Summary

Constructor Summary

Method Summary

Methods inherited from class java.lang.Object

Field Detail

balance_classes

class_sampling_factors

max_after_balance_size

max_confusion_matrix_size

max_hit_ratio_k

overwrite_with_best_model

autoencoder

use_all_factor_levels

activation

hidden

epochs

train_samples_per_iteration

target_ratio_comm_to_comp

seed

adaptive_rate

rho

epsilon

rate

rate_annealing

rate_decay

momentum_start

momentum_ramp

momentum_stable

nesterov_accelerated_gradient

input_dropout_ratio

hidden_dropout_ratios

l1

l2

max_w2

initial_weight_distribution

initial_weight_scale

loss

distribution

tweedie_power

score_interval

score_training_samples

score_validation_samples

score_duty_cycle

classification_stop

regression_stop

quiet_mode

score_validation_sampling

diagnostics

variable_importances

fast_mode

force_load_balance

replicate_training_data

single_node_mode

shuffle_training_data

missing_values_handling

sparse

col_major

average_activation

sparsity_beta

max_categorical_features

reproducible

export_weights_and_biases

elastic_averaging

elastic_averaging_moving_rate

elastic_averaging_regularization

model_id

training_frame

validation_frame

nfolds

keep_cross_validation_predictions

response_column

weights_column

offset_column

fold_column

fold_assignment

ignored_columns

ignore_const_cols

score_each_iteration

checkpoint