water.bindings.pojos

Class SharedTreeParametersV3

java.lang.Object

water.bindings.pojos.Schema

water.bindings.pojos.ModelParametersSchema

water.bindings.pojos.SharedTreeParametersV3

Direct Known Subclasses:

DRFParametersV3, GBMParametersV3

public class SharedTreeParametersV3
extends ModelParametersSchema

Field Summary

Fields

Modifier and Type	Field and Description
boolean	balance_classes Balance training data class counts via over/under-sampling (for imbalanced data).
boolean	build_tree_one_node Run on one node only; no network overhead but fewer cpus used.
float[]	class_sampling_factors Desired over/under-sampling ratios per class (in lexicographic order).
double	col_sample_rate_change_per_level Relative change of the column sampling rate for every level (from 0.0 to 2.0)
double	col_sample_rate_per_tree Column sample rate per tree (from 0.0 to 1.0)
TreeSharedTreeModelSharedTreeParametersHistogramType	histogram_type What type of histogram to use for finding optimal split points
float	max_after_balance_size Maximum relative size of the training data after balancing class counts (can be less than 1.0).
int	max_confusion_matrix_size Maximum size (# classes) for confusion matrices to be printed in the Logs
int	max_depth Maximum tree depth.
int	max_hit_ratio_k Max.
double	min_rows Fewest allowed (weighted) observations in a leaf (in R called 'nodesize').
double	min_split_improvement Minimum relative improvement in squared error reduction for a split to happen
int	nbins For numerical columns (real/int), build a histogram of (at least) this many bins, then split at the best point
int	nbins_cats For categorical columns (factors), build a histogram of this many bins, then split at the best point.
int	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
int	ntrees Number of trees.
double	r2_stopping Stop making trees when the R^2 metric equals or exceeds this
double	sample_rate Row sample rate per tree (from 0.0 to 1.0)
double[]	sample_rate_per_class Row sample rate per tree per class (from 0.0 to 1.0)
int	score_tree_interval Score the model after every so many trees.
long	seed Seed for pseudo random number generator (if applicable)

Fields inherited from class water.bindings.pojos.ModelParametersSchema

checkpoint, fold_assignment, fold_column, ignore_const_cols, ignored_columns, keep_cross_validation_fold_assignment, keep_cross_validation_predictions, max_runtime_secs, model_id, nfolds, offset_column, parallelize_cross_validation, response_column, score_each_iteration, stopping_metric, stopping_rounds, stopping_tolerance, training_frame, validation_frame, weights_column

Constructor Summary

Constructors

Constructor and Description
SharedTreeParametersV3()

Method Summary

Methods

Modifier and Type	Method and Description
java.lang.String	toString() Return the contents of this object as a JSON String.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, 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)

ntrees

public int ntrees

Number of trees.

max_depth

public int max_depth

Maximum tree depth.

min_rows

public double min_rows

Fewest allowed (weighted) observations in a leaf (in R called 'nodesize').

nbins

public int nbins

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

nbins_top_level

public int 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

nbins_cats

public int 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.

r2_stopping

public double r2_stopping

Stop making trees when the R^2 metric equals or exceeds this

seed

public long seed

Seed for pseudo random number generator (if applicable)

build_tree_one_node

public boolean build_tree_one_node

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

sample_rate

public double sample_rate

Row sample rate per tree (from 0.0 to 1.0)

sample_rate_per_class

public double[] sample_rate_per_class

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

col_sample_rate_per_tree

public double col_sample_rate_per_tree

Column sample rate per tree (from 0.0 to 1.0)

col_sample_rate_change_per_level

public double col_sample_rate_change_per_level

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

score_tree_interval

public int score_tree_interval

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

min_split_improvement

public double min_split_improvement

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

histogram_type

public TreeSharedTreeModelSharedTreeParametersHistogramType histogram_type

What type of histogram to use for finding optimal split points

Constructor Detail

SharedTreeParametersV3

public SharedTreeParametersV3()

Method Detail

toString

public java.lang.String toString()

Return the contents of this object as a JSON String.

Overrides:

toString in class ModelParametersSchema