org.apache.ignite.ml.selection.cv

Class AbstractCrossValidation<M extends IgniteModel<Vector,L>,L,K,V>

java.lang.Object

org.apache.ignite.ml.selection.cv.AbstractCrossValidation<M,L,K,V>

Type Parameters:

M - Type of model.

L - Type of a label (truth or prediction).

K - Type of a key in upstream data.

V - Type of a value in upstream data.

Direct Known Subclasses:

CrossValidation, DebugCrossValidation

public abstract class AbstractCrossValidation<M extends IgniteModel<Vector,L>,L,K,V>
extends Object

Cross validation score calculator. Cross validation is an approach that allows to avoid overfitting that is made the following way: the training set is split into k smaller sets. The following procedure is followed for each of the k “folds”:

• A model is trained using k-1 of the folds as training data;
• the resulting model is validated on the remaining part of the data (i.e., it is used as a test set to compute a performance measure such as accuracy).

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	AbstractCrossValidation.TaskResult Represents the scores and map of parameters.

Field Summary

Fields

Modifier and Type	Field and Description
protected int	amountOfFolds Amount of folds.
protected LearningEnvironmentBuilder	envBuilder Learning environment builder.
protected LearningEnvironment	environment Learning Environment.
protected IgniteBiPredicate<K,V>	filter Filter.
protected boolean	isRunningOnPipeline Execution over the pipeline or the chain of preprocessors and separate trainer, otherwise.
protected UniformMapper<K,V>	mapper Mapper.
protected Metric<L>	metric Metric.
protected ParamGrid	paramGrid Parameter grid.
protected int	parts Parts.
protected Pipeline<K,V,Integer,Double>	pipeline Pipeline.
protected Preprocessor<K,V>	preprocessor Preprocessor.
protected DatasetTrainer<M,L>	trainer Trainer.

Constructor Summary

Constructors

Constructor and Description
AbstractCrossValidation()

Method Summary

Modifier and Type	Method and Description
AbstractCrossValidation<M,L,K,V>	isRunningOnPipeline(boolean runningOnPipeline)
protected double[]	score(Function<IgniteBiPredicate<K,V>,DatasetBuilder<K,V>> datasetBuilderSupplier, BiFunction<IgniteBiPredicate<K,V>,M,LabelPairCursor<L>> testDataIterSupplier) Computes cross-validated metrics.
abstract double[]	scoreByFolds() Calculates score by folds.
protected double[]	scorePipeline(Function<IgniteBiPredicate<K,V>,DatasetBuilder<K,V>> datasetBuilderSupplier, BiFunction<IgniteBiPredicate<K,V>,M,LabelPairCursor<L>> testDataIterSupplier) Computes cross-validated metrics.
CrossValidationResult	tuneHyperParamterers() Finds the best set of hyperparameters based on parameter search strategy.
AbstractCrossValidation<M,L,K,V>	withAmountOfFolds(int amountOfFolds)
AbstractCrossValidation<M,L,K,V>	withEnvironmentBuilder(LearningEnvironmentBuilder envBuilder) Changes learning Environment.
AbstractCrossValidation<M,L,K,V>	withFilter(IgniteBiPredicate<K,V> filter)
AbstractCrossValidation<M,L,K,V>	withMapper(UniformMapper<K,V> mapper)
AbstractCrossValidation<M,L,K,V>	withMetric(Metric<L> metric)
AbstractCrossValidation<M,L,K,V>	withParamGrid(ParamGrid paramGrid)
AbstractCrossValidation<M,L,K,V>	withPipeline(Pipeline<K,V,Integer,Double> pipeline)
AbstractCrossValidation<M,L,K,V>	withPreprocessor(Preprocessor<K,V> preprocessor)
AbstractCrossValidation<M,L,K,V>	withTrainer(DatasetTrainer<M,L> trainer)

Methods inherited from class java.lang.Object

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

Field Detail

envBuilder

protected LearningEnvironmentBuilder envBuilder

Learning environment builder.

environment

protected LearningEnvironment environment

Learning Environment.

trainer

protected DatasetTrainer<M extends IgniteModel<Vector,L>,L> trainer

Trainer.

pipeline

protected Pipeline<K,V,Integer,Double> pipeline

Pipeline.

metric

protected Metric<L> metric

Metric.

preprocessor

protected Preprocessor<K,V> preprocessor

Preprocessor.

filter

protected IgniteBiPredicate<K,V> filter

Filter.

amountOfFolds

protected int amountOfFolds

Amount of folds.

parts

protected int parts

Parts.

paramGrid

protected ParamGrid paramGrid

Parameter grid.

isRunningOnPipeline

protected boolean isRunningOnPipeline

Execution over the pipeline or the chain of preprocessors and separate trainer, otherwise.

mapper

protected UniformMapper<K,V> mapper

Mapper.

Constructor Detail

AbstractCrossValidation

public AbstractCrossValidation()

Method Detail

tuneHyperParamterers

public CrossValidationResult tuneHyperParamterers()

Finds the best set of hyperparameters based on parameter search strategy.

scoreByFolds

public abstract double[] scoreByFolds()

Calculates score by folds.

score

protected double[] score(Function<IgniteBiPredicate<K,V>,DatasetBuilder<K,V>> datasetBuilderSupplier,
                         BiFunction<IgniteBiPredicate<K,V>,M,LabelPairCursor<L>> testDataIterSupplier)

Computes cross-validated metrics.

Parameters:

datasetBuilderSupplier - Dataset builder supplier.

testDataIterSupplier - Test data iterator supplier.

Returns:

Array of scores of the estimator for each run of the cross validation.

scorePipeline

protected double[] scorePipeline(Function<IgniteBiPredicate<K,V>,DatasetBuilder<K,V>> datasetBuilderSupplier,
                                 BiFunction<IgniteBiPredicate<K,V>,M,LabelPairCursor<L>> testDataIterSupplier)

Computes cross-validated metrics.

Parameters:

datasetBuilderSupplier - Dataset builder supplier.

testDataIterSupplier - Test data iterator supplier.

Returns:

Array of scores of the estimator for each run of the cross validation.

withTrainer

public AbstractCrossValidation<M,L,K,V> withTrainer(DatasetTrainer<M,L> trainer)

Parameters:

trainer - Trainer.

withMetric

public AbstractCrossValidation<M,L,K,V> withMetric(Metric<L> metric)

Parameters:

metric - Metric.

withPreprocessor

public AbstractCrossValidation<M,L,K,V> withPreprocessor(Preprocessor<K,V> preprocessor)

Parameters:

preprocessor - Preprocessor.

withFilter

public AbstractCrossValidation<M,L,K,V> withFilter(IgniteBiPredicate<K,V> filter)

Parameters:

filter - Filter.

withAmountOfFolds

public AbstractCrossValidation<M,L,K,V> withAmountOfFolds(int amountOfFolds)

Parameters:

amountOfFolds - Amount of folds.

withParamGrid

public AbstractCrossValidation<M,L,K,V> withParamGrid(ParamGrid paramGrid)

Parameters:

paramGrid - Parameter grid.

isRunningOnPipeline

public AbstractCrossValidation<M,L,K,V> isRunningOnPipeline(boolean runningOnPipeline)

Parameters:

runningOnPipeline - Running on pipeline.

withEnvironmentBuilder

public AbstractCrossValidation<M,L,K,V> withEnvironmentBuilder(LearningEnvironmentBuilder envBuilder)

Changes learning Environment.

Parameters:

envBuilder - Learning environment builder.

withPipeline

public AbstractCrossValidation<M,L,K,V> withPipeline(Pipeline<K,V,Integer,Double> pipeline)

Parameters:

pipeline - Pipeline.

withMapper

public AbstractCrossValidation<M,L,K,V> withMapper(UniformMapper<K,V> mapper)

Parameters:

mapper - Mapper.