org.apache.ignite.ml.tree.randomforest

Class RandomForestTrainer<L,S extends ImpurityComputer<BootstrappedVector,S>,T extends RandomForestTrainer<L,S,T>>

java.lang.Object

org.apache.ignite.ml.trainers.DatasetTrainer<M,Double>

org.apache.ignite.ml.trainers.SingleLabelDatasetTrainer<ModelsComposition>

org.apache.ignite.ml.tree.randomforest.RandomForestTrainer<L,S,T>

Type Parameters:

L - Type of statistics aggregator for leaf values computing.

S - Type of impurity computer specific for algorithm.

T - Type of child of RandomForestTrainer using in with-methods.

Direct Known Subclasses:

RandomForestClassifierTrainer, RandomForestRegressionTrainer

public abstract class RandomForestTrainer<L,S extends ImpurityComputer<BootstrappedVector,S>,T extends RandomForestTrainer<L,S,T>>
extends SingleLabelDatasetTrainer<ModelsComposition>

Class represents a realization of Random Forest algorithm. Main idea of this realization is that at each learning iteration it tries to aggregate statistics on impurity for each corner nodes (leaves that may be splitted on a new iteration) in each trees in random forest. It requires one map-reduce operation over learning dataset. After such aggregation the algorithm selects split points for each corner nodes of create leaf nodes. The algorithm stops when there is no splitting for nodes in trees. At last stage the algorithm aggregates statistics on labels for leaf nodes in one map-reduce step and sets values to leafs based these statistics.

Nested Class Summary

Nested classes/interfaces inherited from class org.apache.ignite.ml.trainers.DatasetTrainer

DatasetTrainer.EmptyDatasetException

Field Summary

Fields inherited from class org.apache.ignite.ml.trainers.DatasetTrainer

envBuilder, environment

Constructor Summary

Constructors

Constructor and Description
RandomForestTrainer(List<FeatureMeta> meta) Create an instance of RandomForestTrainer.

Method Summary

Modifier and Type	Method and Description
protected abstract ModelsComposition	buildComposition(List<TreeRoot> models) Returns composition of built trees.
protected abstract ImpurityHistogramsComputer<S>	createImpurityHistogramsComputer() Creates an instance of Histograms Computer corresponding to RF implementation.
protected abstract LeafValuesComputer<L>	createLeafStatisticsAggregator() Creates an instance of Leaf Statistics Aggregator corresponding to RF implementation.
<K,V> ModelsComposition	fitWithInitializedDeployingContext(DatasetBuilder<K,V> datasetBuilder, Preprocessor<K,V> preprocessor) Trains model based on the specified data.
protected boolean	init(Dataset<EmptyContext,BootstrappedDatasetPartition> dataset) Init-step before learning.
protected ArrayList<TreeRoot>	initTrees(Queue<TreeNode> treesQueue) Creates list of trees.
protected abstract T	instance()
boolean	isUpdateable(ModelsComposition mdl)
protected <K,V> ModelsComposition	updateModel(ModelsComposition mdl, DatasetBuilder<K,V> datasetBuilder, Preprocessor<K,V> preprocessor) Trains new model taken previous one as a first approximation.
T	withAmountOfTrees(int amountOfTrees)
T	withFeaturesCountSelectionStrgy(Function<List<FeatureMeta>,Integer> strgy)
T	withMaxDepth(int maxDepth)
T	withMinImpurityDelta(double minImpurityDelta)
T	withNodesToLearnSelectionStrgy(Function<Queue<TreeNode>,List<TreeNode>> strgy) Sets strategy for selection nodes from learning queue in each iteration.
T	withSeed(long seed)
T	withSubSampleSize(double subSampleSize)

Methods inherited from class org.apache.ignite.ml.trainers.DatasetTrainer

fit, fit, fit, fit, fit, fit, getLastTrainedModelOrThrowEmptyDatasetException, identityTrainer, learningEnvironment, update, update, update, update, update, withConvertedLabels, withEnvironmentBuilder

Methods inherited from class java.lang.Object

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

Constructor Detail

RandomForestTrainer

public RandomForestTrainer(List<FeatureMeta> meta)

Create an instance of RandomForestTrainer.

Parameters:

meta - Features Meta.

Method Detail

fitWithInitializedDeployingContext

public <K,V> ModelsComposition fitWithInitializedDeployingContext(DatasetBuilder<K,V> datasetBuilder,
                                                                  Preprocessor<K,V> preprocessor)

Trains model based on the specified data.

Specified by:

fitWithInitializedDeployingContext in class DatasetTrainer<ModelsComposition,Double>

Type Parameters:

K - Type of a key in upstream data.

V - Type of a value in upstream data.

Parameters:

datasetBuilder - Dataset builder.

preprocessor - Extractor of UpstreamEntry into LabeledVector.

Returns:

Model.

instance

protected abstract T instance()

Returns:

an instance of current object with valid type in according to inheritance.

withAmountOfTrees

public T withAmountOfTrees(int amountOfTrees)

Parameters:

amountOfTrees - Count of trees.

Returns:

an instance of current object with valid type in according to inheritance.

withSubSampleSize

public T withSubSampleSize(double subSampleSize)

Parameters:

subSampleSize - Subsample size.

Returns:

an instance of current object with valid type in according to inheritance.

withMaxDepth

public T withMaxDepth(int maxDepth)

Parameters:

maxDepth - Max depth.

Returns:

an instance of current object with valid type in according to inheritance.

withMinImpurityDelta

public T withMinImpurityDelta(double minImpurityDelta)

Parameters:

minImpurityDelta - Min impurity delta.

Returns:

an instance of current object with valid type in according to inheritance.

withFeaturesCountSelectionStrgy

public T withFeaturesCountSelectionStrgy(Function<List<FeatureMeta>,Integer> strgy)

Parameters:

strgy - Strgy.

Returns:

an instance of current object with valid type in according to inheritance.

withNodesToLearnSelectionStrgy

public T withNodesToLearnSelectionStrgy(Function<Queue<TreeNode>,List<TreeNode>> strgy)

Sets strategy for selection nodes from learning queue in each iteration.

Parameters:

strgy - Strgy.

withSeed

public T withSeed(long seed)

Parameters:

seed - Seed.

Returns:

an instance of current object with valid type in according to inheritance.

init

protected boolean init(Dataset<EmptyContext,BootstrappedDatasetPartition> dataset)

Init-step before learning. It may be useful collecting labels statistics step for classification.

Parameters:

dataset - Dataset.

isUpdateable

public boolean isUpdateable(ModelsComposition mdl)

Specified by:

isUpdateable in class DatasetTrainer<ModelsComposition,Double>

Parameters:

mdl - Model.

Returns:

True if current critical for training parameters correspond to parameters from last training.

updateModel

protected <K,V> ModelsComposition updateModel(ModelsComposition mdl,
                                              DatasetBuilder<K,V> datasetBuilder,
                                              Preprocessor<K,V> preprocessor)

Trains new model taken previous one as a first approximation.

Specified by:

updateModel in class DatasetTrainer<ModelsComposition,Double>

Type Parameters:

K - Type of a key in upstream data.

V - Type of a value in upstream data.

Parameters:

mdl - Learned model.

datasetBuilder - Dataset builder.

preprocessor - Extractor of UpstreamEntry into LabeledVector.

Returns:

Updated model.

createImpurityHistogramsComputer

protected abstract ImpurityHistogramsComputer<S> createImpurityHistogramsComputer()

Creates an instance of Histograms Computer corresponding to RF implementation.

createLeafStatisticsAggregator

protected abstract LeafValuesComputer<L> createLeafStatisticsAggregator()

Creates an instance of Leaf Statistics Aggregator corresponding to RF implementation.

initTrees

protected ArrayList<TreeRoot> initTrees(Queue<TreeNode> treesQueue)

Creates list of trees.

Parameters:

treesQueue - Trees queue.

Returns:

List of trees.

buildComposition

protected abstract ModelsComposition buildComposition(List<TreeRoot> models)

Returns composition of built trees.

Parameters:

models - Models.

Returns:

composition of built trees.