org.apache.ignite.ml.trainers

Class AdaptableDatasetTrainer<I,O,IW,OW,M extends IgniteModel<IW,OW>,L>

java.lang.Object

org.apache.ignite.ml.trainers.DatasetTrainer<AdaptableDatasetModel<I,O,IW,OW,M>,L>

org.apache.ignite.ml.trainers.AdaptableDatasetTrainer<I,O,IW,OW,M,L>

Type Parameters:

I - Input type of model produced by this trainer.

O - Output type of model produced by this trainer.

IW - Input type of model produced by wrapped trainer.

OW - Output type of model produced by wrapped trainer.

M - Type of model produced by wrapped model.

L - Type of labels.

public class AdaptableDatasetTrainer<I,O,IW,OW,M extends IgniteModel<IW,OW>,L>
extends DatasetTrainer<AdaptableDatasetModel<I,O,IW,OW,M>,L>

Type used to adapt input and output types of wrapped DatasetTrainer. Produces model which is composition of form before `andThen` wMdl `andThen` after where wMdl is model produced by wrapped trainer.

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

Method Summary

Modifier and Type	Method and Description
AdaptableDatasetTrainer<I,O,IW,OW,M,L>	afterFeatureExtractor(IgniteFunction<Vector,Vector> after) Specify function which will be applied after feature extractor.
AdaptableDatasetTrainer<I,O,IW,OW,M,L>	afterLabelExtractor(IgniteFunction<L,L> after) Specify function which will be applied after label extractor.
<O1> AdaptableDatasetTrainer<I,O1,IW,OW,M,L>	afterTrainedModel(IgniteFunction<O,O1> after) Let this trainer produce model mdl.
<O1,M1 extends IgniteModel<O,O1>> TrainersSequentialComposition<I,O,O1,L>	andThen(DatasetTrainer<M1,L> tr, IgniteFunction<AdaptableDatasetModel<I,O,IW,OW,M>,IgniteFunction<LabeledVector<L>,LabeledVector<L>>> datasetMappingProducer) Create a TrainersSequentialComposition of whis trainer and specified trainer.
<I1> AdaptableDatasetTrainer<I1,O,IW,OW,M,L>	beforeTrainedModel(IgniteFunction<I1,I> before) Let this trainer produce model mdl.
<K,V> AdaptableDatasetModel<I,O,IW,OW,M>	fitWithInitializedDeployingContext(DatasetBuilder<K,V> datasetBuilder, Preprocessor<K,V> extractor) Trains model based on the specified data.
boolean	isUpdateable(AdaptableDatasetModel<I,O,IW,OW,M> mdl)
static <I,O,M extends IgniteModel<I,O>,L> AdaptableDatasetTrainer<I,O,I,O,M,L>	of(DatasetTrainer<M,L> wrapped) Construct instance of this class from a given DatasetTrainer.
protected <K,V> AdaptableDatasetModel<I,O,IW,OW,M>	updateModel(AdaptableDatasetModel<I,O,IW,OW,M> mdl, DatasetBuilder<K,V> datasetBuilder, Preprocessor<K,V> extractor) Trains new model taken previous one as a first approximation.
AdaptableDatasetTrainer<I,O,IW,OW,M,L>	withDatasetMapping(DatasetMapping<L,L> mapping) Specify DatasetMapping which will be applied to dataset before fitting and updating.
AdaptableDatasetTrainer<I,O,IW,OW,M,L>	withUpstreamTransformerBuilder(UpstreamTransformerBuilder upstreamTransformerBuilder) Specify which UpstreamTransformerBuilder will be used.

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

Method Detail

of

public static <I,O,M extends IgniteModel<I,O>,L> AdaptableDatasetTrainer<I,O,I,O,M,L> of(DatasetTrainer<M,L> wrapped)

Construct instance of this class from a given DatasetTrainer.

Type Parameters:

I - Input type of wrapped trainer.

O - Output type of wrapped trainer.

M - Type of model produced by wrapped trainer.

L - Type of labels.

Parameters:

wrapped - Wrapped trainer.

Returns:

Instance of this class.

fitWithInitializedDeployingContext

public <K,V> AdaptableDatasetModel<I,O,IW,OW,M> fitWithInitializedDeployingContext(DatasetBuilder<K,V> datasetBuilder,
                                                                                   Preprocessor<K,V> extractor)

Trains model based on the specified data.

Specified by:

fitWithInitializedDeployingContext in class DatasetTrainer<AdaptableDatasetModel<I,O,IW,OW,M extends IgniteModel<IW,OW>>,L>

Type Parameters:

K - Type of a key in upstream data.

V - Type of a value in upstream data.

Parameters:

datasetBuilder - Dataset builder.

extractor - Extractor of UpstreamEntry into LabeledVector.

Returns:

Model.

isUpdateable

public boolean isUpdateable(AdaptableDatasetModel<I,O,IW,OW,M> mdl)

Specified by:

isUpdateable in class DatasetTrainer<AdaptableDatasetModel<I,O,IW,OW,M extends IgniteModel<IW,OW>>,L>

Parameters:

mdl - Model.

Returns:

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

updateModel

protected <K,V> AdaptableDatasetModel<I,O,IW,OW,M> updateModel(AdaptableDatasetModel<I,O,IW,OW,M> mdl,
                                                               DatasetBuilder<K,V> datasetBuilder,
                                                               Preprocessor<K,V> extractor)

Trains new model taken previous one as a first approximation.

Specified by:

updateModel in class DatasetTrainer<AdaptableDatasetModel<I,O,IW,OW,M extends IgniteModel<IW,OW>>,L>

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.

extractor - Extractor of UpstreamEntry into LabeledVector.

Returns:

Updated model.

afterTrainedModel

public <O1> AdaptableDatasetTrainer<I,O1,IW,OW,M,L> afterTrainedModel(IgniteFunction<O,O1> after)

Let this trainer produce model mdl. This method produces a trainer which produces mdl1, where mdl1 = mdl `andThen` after.

Type Parameters:

O1 - Type of produced model output.

Parameters:

after - Function inserted before produced model.

Returns:

New DatasetTrainer which produces composition of specified function and model produced by original trainer.

beforeTrainedModel

public <I1> AdaptableDatasetTrainer<I1,O,IW,OW,M,L> beforeTrainedModel(IgniteFunction<I1,I> before)

Let this trainer produce model mdl. This method produces a trainer which produces mdl1, where mdl1 = f `andThen` mdl.

Type Parameters:

I1 - Type of produced model input.

Parameters:

before - Function inserted before produced model.

Returns:

New DatasetTrainer which produces composition of specified function and model produced by original trainer.

withDatasetMapping

public AdaptableDatasetTrainer<I,O,IW,OW,M,L> withDatasetMapping(DatasetMapping<L,L> mapping)

Specify DatasetMapping which will be applied to dataset before fitting and updating.

Parameters:

mapping - DatasetMapping which will be applied to dataset before fitting and updating.

Returns:

New trainer of the same type, but with specified mapping applied to dataset before fitting and updating.

andThen

public <O1,M1 extends IgniteModel<O,O1>> TrainersSequentialComposition<I,O,O1,L> andThen(DatasetTrainer<M1,L> tr,
                                                                                         IgniteFunction<AdaptableDatasetModel<I,O,IW,OW,M>,IgniteFunction<LabeledVector<L>,LabeledVector<L>>> datasetMappingProducer)

Create a TrainersSequentialComposition of whis trainer and specified trainer.

Type Parameters:

O1 - Type of output of trainer to compose with.

M1 - Type of model produced by the trainer to compose with.

Parameters:

tr - Trainer to compose with.

datasetMappingProducer - DatasetMapping producer specifying dependency between this trainer and trainer to compose with.

Returns:

A TrainersSequentialComposition of whis trainer and specified trainer.

afterFeatureExtractor

public AdaptableDatasetTrainer<I,O,IW,OW,M,L> afterFeatureExtractor(IgniteFunction<Vector,Vector> after)

Specify function which will be applied after feature extractor.

Parameters:

after - Function which will be applied after feature extractor.

Returns:

New trainer with same parameters as this trainer except that specified function will be applied after feature extractor.

afterLabelExtractor

public AdaptableDatasetTrainer<I,O,IW,OW,M,L> afterLabelExtractor(IgniteFunction<L,L> after)

Specify function which will be applied after label extractor.

Parameters:

after - Function which will be applied after label extractor.

Returns:

New trainer with same parameters as this trainer has except that specified function will be applied after label extractor.

withUpstreamTransformerBuilder

public AdaptableDatasetTrainer<I,O,IW,OW,M,L> withUpstreamTransformerBuilder(UpstreamTransformerBuilder upstreamTransformerBuilder)

Specify which UpstreamTransformerBuilder will be used.

Parameters:

upstreamTransformerBuilder - UpstreamTransformerBuilder to use.

Returns:

New trainer with same parameters as this trainer has except that specified UpstreamTransformerBuilder will be used.