org.apache.ignite.ml.composition.combinators.sequential

Class TrainersSequentialComposition<I,O1,O2,L>

java.lang.Object

org.apache.ignite.ml.trainers.DatasetTrainer<ModelsSequentialComposition<I,O1,O2>,L>

org.apache.ignite.ml.composition.combinators.sequential.TrainersSequentialComposition<I,O1,O2,L>

Type Parameters:

I - Type of input of model produced by first trainer.

O1 - Type of output of model produced by first trainer.

O2 - Type of output of model produced by second trainer.

L - Type of labels.

public class TrainersSequentialComposition<I,O1,O2,L>
extends DatasetTrainer<ModelsSequentialComposition<I,O1,O2>,L>

Sequential composition of trainers. Sequential composition of trainers is itself trainer which produces ModelsSequentialComposition. Training is done in following fashion:

     1. First trainer is trained and `mdl1` is produced.
     2. From `mdl1` DatasetMapping is constructed. This mapping `dsM` encapsulates dependency between first
     training result and second trainer.
     3. Second trainer is trained using dataset aquired from application `dsM` to original dataset; `mdl2` is produced.
     4. `mdl1` and `mdl2` are composed into ModelsSequentialComposition.
 

Nested Class Summary

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

DatasetTrainer.EmptyDatasetException

Field Summary

Fields

Modifier and Type	Field and Description
protected IgniteBiFunction<Integer,? super IgniteModel<I,O1>,IgniteFunction<LabeledVector<L>,LabeledVector<L>>>	datasetMapping Dataset mapping.

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

envBuilder, environment

Constructor Summary

Constructors

Constructor and Description
TrainersSequentialComposition(DatasetTrainer<? extends IgniteModel<I,O1>,L> tr1, DatasetTrainer<? extends IgniteModel<O1,O2>,L> tr2, IgniteBiFunction<Integer,? super IgniteModel<I,O1>,IgniteFunction<LabeledVector<L>,LabeledVector<L>>> datasetMapping) Create sequential composition of two trainers.
TrainersSequentialComposition(DatasetTrainer<? extends IgniteModel<I,O1>,L> tr1, DatasetTrainer<? extends IgniteModel<O1,O2>,L> tr2, IgniteFunction<? super IgniteModel<I,O1>,IgniteFunction<LabeledVector<L>,LabeledVector<L>>> datasetMapping) Construct sequential composition of given two trainers.

Method Summary

Modifier and Type	Method and Description
<K,V> ModelsSequentialComposition<I,O1,O2>	fitWithInitializedDeployingContext(DatasetBuilder<K,V> datasetBuilder, Preprocessor<K,V> preprocessor) Trains model based on the specified data.
boolean	isUpdateable(ModelsSequentialComposition<I,O1,O2> mdl) This method is never called, instead of constructing logic of update from DatasetTrainer.isUpdateable(M) and DatasetTrainer.updateModel(M, org.apache.ignite.ml.dataset.DatasetBuilder<K, V>, org.apache.ignite.ml.preprocessing.Preprocessor<K, V>) in this class we explicitly override update method.
static <I,O,L> TrainersSequentialComposition<I,O,O,L>	ofSame(DatasetTrainer<? extends IgniteModel<I,O>,L> tr, IgniteBiFunction<Integer,? super IgniteModel<I,O>,IgniteFunction<LabeledVector<L>,LabeledVector<L>>> datasetMapping, IgniteBiPredicate<Integer,IgniteModel<I,O>> shouldStop, IgniteFunction<O,I> out2In) Construct sequential composition of same trainers.
DatasetTrainer<IgniteModel<I,O2>,L>	unsafeSimplyTyped() Performs coersion of this trainer to DatasetTrainer<IgniteModel<I, O2>, L>.
<K,V> ModelsSequentialComposition<I,O1,O2>	update(ModelsSequentialComposition<I,O1,O2> mdl, DatasetBuilder<K,V> datasetBuilder, Preprocessor<K,V> preprocessor) Gets state of model in arguments, compare it with training parameters of trainer and if they are fit then trainer updates model in according to new data and return new model.
protected <K,V> ModelsSequentialComposition<I,O1,O2>	updateModel(ModelsSequentialComposition<I,O1,O2> mdl, DatasetBuilder<K,V> datasetBuilder, Preprocessor<K,V> preprocessor) This method is never called, instead of constructing logic of update from DatasetTrainer.isUpdateable(IgniteModel) and DatasetTrainer.updateModel(IgniteModel, DatasetBuilder, Preprocessor) in this class we explicitly override update method.

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

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

Methods inherited from class java.lang.Object

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

Field Detail

datasetMapping

protected IgniteBiFunction<Integer,? super IgniteModel<I,O1>,IgniteFunction<LabeledVector<L>,LabeledVector<L>>> datasetMapping

Dataset mapping.

Constructor Detail

TrainersSequentialComposition

public TrainersSequentialComposition(DatasetTrainer<? extends IgniteModel<I,O1>,L> tr1,
                                     DatasetTrainer<? extends IgniteModel<O1,O2>,L> tr2,
                                     IgniteFunction<? super IgniteModel<I,O1>,IgniteFunction<LabeledVector<L>,LabeledVector<L>>> datasetMapping)

Construct sequential composition of given two trainers.

Parameters:

tr1 - First trainer.

tr2 - Second trainer.

datasetMapping - Dataset mapping.

TrainersSequentialComposition

public TrainersSequentialComposition(DatasetTrainer<? extends IgniteModel<I,O1>,L> tr1,
                                     DatasetTrainer<? extends IgniteModel<O1,O2>,L> tr2,
                                     IgniteBiFunction<Integer,? super IgniteModel<I,O1>,IgniteFunction<LabeledVector<L>,LabeledVector<L>>> datasetMapping)

Create sequential composition of two trainers.

Parameters:

tr1 - First trainer.

tr2 - Second trainer.

datasetMapping - Dataset mapping containing dependence between first and second trainer.

Method Detail

ofSame

public static <I,O,L> TrainersSequentialComposition<I,O,O,L> ofSame(DatasetTrainer<? extends IgniteModel<I,O>,L> tr,
                                                                    IgniteBiFunction<Integer,? super IgniteModel<I,O>,IgniteFunction<LabeledVector<L>,LabeledVector<L>>> datasetMapping,
                                                                    IgniteBiPredicate<Integer,IgniteModel<I,O>> shouldStop,
                                                                    IgniteFunction<O,I> out2In)

Construct sequential composition of same trainers.

Type Parameters:

I - Type of input of model produced by trainer.

O - Type of output of model produced by trainer.

L - Type of labels for trainer.

Parameters:

tr - Trainer used for sequential composition.

datasetMapping - Dataset mapping.

shouldStop - Predicate depending on index and model produced by the last trainer indicating if composition process should stop

out2In - Function for conversion of output of model into input of next model.

Returns:

Sequential composition of same trainers.

fitWithInitializedDeployingContext

public <K,V> ModelsSequentialComposition<I,O1,O2> fitWithInitializedDeployingContext(DatasetBuilder<K,V> datasetBuilder,
                                                                                     Preprocessor<K,V> preprocessor)

Trains model based on the specified data.

Specified by:

fitWithInitializedDeployingContext in class DatasetTrainer<ModelsSequentialComposition<I,O1,O2>,L>

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.

updateModel

protected <K,V> ModelsSequentialComposition<I,O1,O2> updateModel(ModelsSequentialComposition<I,O1,O2> mdl,
                                                                 DatasetBuilder<K,V> datasetBuilder,
                                                                 Preprocessor<K,V> preprocessor)

This method is never called, instead of constructing logic of update from DatasetTrainer.isUpdateable(IgniteModel) and DatasetTrainer.updateModel(IgniteModel, DatasetBuilder, Preprocessor) in this class we explicitly override update method.

Specified by:

updateModel in class DatasetTrainer<ModelsSequentialComposition<I,O1,O2>,L>

Type Parameters:

K - Type of a key in upstream data.

V - Type of a value in upstream data.

Parameters:

mdl - Model.

datasetBuilder - Dataset builder.

preprocessor - Extractor of UpstreamEntry into LabeledVector.

Returns:

Updated model.

update

public <K,V> ModelsSequentialComposition<I,O1,O2> update(ModelsSequentialComposition<I,O1,O2> mdl,
                                                         DatasetBuilder<K,V> datasetBuilder,
                                                         Preprocessor<K,V> preprocessor)

Gets state of model in arguments, compare it with training parameters of trainer and if they are fit then trainer updates model in according to new data and return new model. In other case trains new model.

Overrides:

update in class DatasetTrainer<ModelsSequentialComposition<I,O1,O2>,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.

preprocessor - Extractor of UpstreamEntry into LabeledVector.

Returns:

Updated model.

isUpdateable

public boolean isUpdateable(ModelsSequentialComposition<I,O1,O2> mdl)

This method is never called, instead of constructing logic of update from DatasetTrainer.isUpdateable(M) and DatasetTrainer.updateModel(M, org.apache.ignite.ml.dataset.DatasetBuilder<K, V>, org.apache.ignite.ml.preprocessing.Preprocessor<K, V>) in this class we explicitly override update method.

Specified by:

isUpdateable in class DatasetTrainer<ModelsSequentialComposition<I,O1,O2>,L>

Parameters:

mdl - Model.

Returns:

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

unsafeSimplyTyped

public DatasetTrainer<IgniteModel<I,O2>,L> unsafeSimplyTyped()

Performs coersion of this trainer to DatasetTrainer<IgniteModel<I, O2>, L>.

Returns:

Trainer coerced to DatasetTrainer<IgniteModel<I, O>, L>.