org.apache.ignite.ml.tree.impurity

Class ImpurityMeasureCalculator<T extends ImpurityMeasure<T>>

java.lang.Object

org.apache.ignite.ml.tree.impurity.ImpurityMeasureCalculator<T>

Type Parameters:

T - Type of impurity measure.

All Implemented Interfaces:

Serializable

Direct Known Subclasses:

GiniImpurityMeasureCalculator, MSEImpurityMeasureCalculator

public abstract class ImpurityMeasureCalculator<T extends ImpurityMeasure<T>>
extends Object
implements Serializable

Base interface for impurity measure calculators that calculates all impurity measures required to find a best split.

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field and Description
protected boolean	useIdx Use index structure instead of using sorting while learning.

Constructor Summary

Constructors

Constructor and Description
ImpurityMeasureCalculator(boolean useIdx) Constructs an instance of ImpurityMeasureCalculator.

Method Summary

Modifier and Type	Method and Description
abstract StepFunction<T>[]	calculate(DecisionTreeData data, TreeFilter filter, int depth) Calculates all impurity measures required required to find a best split and returns them as an array of StepFunction (for every column).
protected int	columnsCount(DecisionTreeData data, TreeDataIndex idx) Returns columns count in current dataset.
protected double	getFeatureValue(DecisionTreeData data, TreeDataIndex idx, int featureId, int k) Returns feature value in according to kth order statistic.
protected Vector	getFeatureValues(DecisionTreeData data, TreeDataIndex idx, int featureId, int k) Returns feature value in according to kth order statistic.
protected double	getLabelValue(DecisionTreeData data, TreeDataIndex idx, int featureId, int k) Returns label value in according to kth order statistic.
protected int	rowsCount(DecisionTreeData data, TreeDataIndex idx) Returns rows count in current dataset.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

useIdx

protected final boolean useIdx

Use index structure instead of using sorting while learning.

Constructor Detail

ImpurityMeasureCalculator

public ImpurityMeasureCalculator(boolean useIdx)

Constructs an instance of ImpurityMeasureCalculator.

Parameters:

useIdx - Use index.

Method Detail

calculate

public abstract StepFunction<T>[] calculate(DecisionTreeData data,
                                            TreeFilter filter,
                                            int depth)

Calculates all impurity measures required required to find a best split and returns them as an array of StepFunction (for every column).

Parameters:

data - Features and labels.

Returns:

Impurity measures as an array of StepFunction (for every column).

columnsCount

protected int columnsCount(DecisionTreeData data,
                           TreeDataIndex idx)

Returns columns count in current dataset.

Parameters:

data - Data.

idx - Index.

Returns:

Columns count in current dataset.

rowsCount

protected int rowsCount(DecisionTreeData data,
                        TreeDataIndex idx)

Returns rows count in current dataset.

Parameters:

data - Data.

idx - Index.

Returns:

rows count in current dataset

getLabelValue

protected double getLabelValue(DecisionTreeData data,
                               TreeDataIndex idx,
                               int featureId,
                               int k)

Returns label value in according to kth order statistic.

Parameters:

data - Data.

idx - Index.

featureId - Feature id.

k - K-th statistic.

Returns:

label value in according to kth order statistic

getFeatureValue

protected double getFeatureValue(DecisionTreeData data,
                                 TreeDataIndex idx,
                                 int featureId,
                                 int k)

Returns feature value in according to kth order statistic.

Parameters:

data - Data.

idx - Index.

featureId - Feature id.

k - K-th statistic.

Returns:

feature value in according to kth order statistic.

getFeatureValues

protected Vector getFeatureValues(DecisionTreeData data,
                                  TreeDataIndex idx,
                                  int featureId,
                                  int k)

Returns feature value in according to kth order statistic.

Parameters:

data - Data.

idx - Index.

featureId - Feature id.

k - K-th statistic.

Returns:

feature value in according to kth order statistic.