Package org.tribuo.ensemble

Class WeightedEnsembleModel<T extends Output<T>>

java.lang.Object

org.tribuo.Model<T>

org.tribuo.ensemble.EnsembleModel<T>

org.tribuo.ensemble.WeightedEnsembleModel<T>

All Implemented Interfaces:

com.oracle.labs.mlrg.olcut.provenance.Provenancable<ModelProvenance>, Serializable, ONNXExportable

public final class WeightedEnsembleModel<T extends Output<T>>
extends EnsembleModel<T>
implements ONNXExportable

An ensemble model that uses weights to combine the ensemble member predictions.

See Also:

• Serialized Form

Field Summary

Fields

Modifier and Type	Field	Description
protected final EnsembleCombiner<T>	combiner
protected final float[]	weights

Fields inherited from class org.tribuo.ensemble.EnsembleModel

models

Fields inherited from class org.tribuo.Model

ALL_OUTPUTS, BIAS_FEATURE, featureIDMap, generatesProbabilities, name, outputIDInfo, provenance, provenanceOutput

Fields inherited from interface org.tribuo.ONNXExportable

PROVENANCE_METADATA_FIELD, SERIALIZER

Constructor Summary

Constructors

Constructor	Description
WeightedEnsembleModel(String name, EnsembleModelProvenance provenance, ImmutableFeatureMap featureIDMap, ImmutableOutputInfo<T> outputIDInfo, List<Model<T>> newModels, EnsembleCombiner<T> combiner)	Unless you are implementing a Trainer you should not use this constructor directly.
WeightedEnsembleModel(String name, EnsembleModelProvenance provenance, ImmutableFeatureMap featureIDMap, ImmutableOutputInfo<T> outputIDInfo, List<Model<T>> newModels, EnsembleCombiner<T> combiner, float[] weights)	Unless you are implementing a Trainer you should not use this constructor directly.

Method Summary

Modifier and Type	Method	Description
protected EnsembleModel<T>	copy(String name, EnsembleModelProvenance newProvenance, List<Model<T>> newModels)	Copies this ensemble model.
static <T extends Output<T>> WeightedEnsembleModel<T>	createEnsembleFromExistingModels(String name, List<Model<T>> models, EnsembleCombiner<T> combiner)	Creates an ensemble from existing models.
static <T extends Output<T>> WeightedEnsembleModel<T>	createEnsembleFromExistingModels(String name, List<Model<T>> models, EnsembleCombiner<T> combiner, float[] weights)	Creates an ensemble from existing models.
OnnxMl.ModelProto	exportONNXModel(String domain, long modelVersion)	Exports this EnsembleModel as an ONNX model.
Optional<Excuse<T>>	getExcuse(Example<T> example)	Generates an excuse for an example.
Prediction<T>	predict(Example<T> example)	Uses the model to predict the output for a single example.
ONNXNode	writeONNXGraph(ONNXRef<?> input)	Writes this Model into OnnxMl.GraphProto.Builder inside the input's ONNXContext.

Methods inherited from class org.tribuo.ensemble.EnsembleModel

copy, getModels, getNumModels, getProvenance, getTopFeatures

Methods inherited from class org.tribuo.Model

castModel, copy, generatesProbabilities, getExcuses, getFeatureIDMap, getName, getOutputIDInfo, innerPredict, predict, predict, setName, toString, validate

Methods inherited from class java.lang.Object

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

Methods inherited from interface org.tribuo.ONNXExportable

saveONNXModel, serializeProvenance

Field Details

weights

protected final float[] weights

combiner

protected final EnsembleCombiner<T extends Output<T>> combiner

Constructor Details

WeightedEnsembleModel

public WeightedEnsembleModel(String name,
 EnsembleModelProvenance provenance,
 ImmutableFeatureMap featureIDMap,
 ImmutableOutputInfo<T> outputIDInfo,
 List<Model<T>> newModels,
 EnsembleCombiner<T> combiner)

Unless you are implementing a Trainer you should not use this constructor directly. Instead use createEnsembleFromExistingModels(String, List, EnsembleCombiner).

Constructs an ensemble model which uses uniform weights.

Parameters:

name - The model name.

provenance - The model provenance.

featureIDMap - The feature domain.

outputIDInfo - The output domain.

newModels - The list of ensemble members.

combiner - The combination function.

WeightedEnsembleModel

public WeightedEnsembleModel(String name,
 EnsembleModelProvenance provenance,
 ImmutableFeatureMap featureIDMap,
 ImmutableOutputInfo<T> outputIDInfo,
 List<Model<T>> newModels,
 EnsembleCombiner<T> combiner,
 float[] weights)

Unless you are implementing a Trainer you should not use this constructor directly. Instead use createEnsembleFromExistingModels(String, List, EnsembleCombiner, float[]).

Constructs an ensemble model which uses uniform weights.

Parameters:

name - The model name.

provenance - The model provenance.

featureIDMap - The feature domain.

outputIDInfo - The output domain.

newModels - The list of ensemble members.

combiner - The combination function.

weights - The model combination weights.

Method Details

predict

public Prediction<T> predict(Example<T> example)

Description copied from class: Model

Uses the model to predict the output for a single example.

predict does not mutate the example.

Throws IllegalArgumentException if the example has no features or no feature overlap with the model.

Specified by:

predict in class Model<T extends Output<T>>

Parameters:

example - the example to predict.

Returns:

the result of the prediction.

getExcuse

public Optional<Excuse<T>> getExcuse(Example<T> example)

Description copied from class: Model

Generates an excuse for an example.

This attempts to explain a classification result. Generating an excuse may be quite an expensive operation.

This excuse either contains per class information or an entry with key Model.ALL_OUTPUTS.

The optional is empty if the model does not provide excuses.

Specified by:

getExcuse in class EnsembleModel<T extends Output<T>>

Parameters:

example - The input example.

Returns:

An optional excuse object. The optional is empty if this model does not provide excuses.

copy

protected EnsembleModel<T> copy(String name,
 EnsembleModelProvenance newProvenance,
 List<Model<T>> newModels)

Description copied from class: EnsembleModel

Copies this ensemble model.

Specified by:

copy in class EnsembleModel<T extends Output<T>>

Parameters:

name - The new name.

newProvenance - The new provenance.

newModels - The new models.

Returns:

A copy of the ensemble model.

createEnsembleFromExistingModels

public static <T extends Output<T>>
WeightedEnsembleModel<T> createEnsembleFromExistingModels(String name,
 List<Model<T>> models,
 EnsembleCombiner<T> combiner)

Creates an ensemble from existing models. The model outputs are combined using uniform weights.

Uses the feature and output domain from the first model as the ensemble model's domains. The individual ensemble members use the domains that they contain.

If the output domains don't cover the same dimensions then it throws IllegalArgumentException.

Type Parameters:

T - The output type.

Parameters:

name - The ensemble name.

models - The ensemble members.

combiner - The combination function.

Returns:

A weighted ensemble model.

createEnsembleFromExistingModels

public static <T extends Output<T>>
WeightedEnsembleModel<T> createEnsembleFromExistingModels(String name,
 List<Model<T>> models,
 EnsembleCombiner<T> combiner,
 float[] weights)

Creates an ensemble from existing models.

Uses the feature and output domain from the first model as the ensemble model's domains. The individual ensemble members use the domains that they contain.

If the output domains don't cover the same dimensions then it throws IllegalArgumentException. If the weights aren't the same length as the models it throws IllegalArgumentException.

Type Parameters:

T - The output type.

Parameters:

name - The ensemble name.

models - The ensemble members.

combiner - The combination function.

weights - The model combination weights.

Returns:

A weighted ensemble model.

exportONNXModel

public OnnxMl.ModelProto exportONNXModel(String domain,
 long modelVersion)

Exports this EnsembleModel as an ONNX model.

Note if the ensemble members or the ensemble combination function do not implement ONNXExportable then this method will throw UnsupportedOperationException.

Specified by:

exportONNXModel in interface ONNXExportable

Parameters:

domain - A reverse-DNS name to namespace the model (e.g., org.tribuo.classification.sgd.linear).

modelVersion - A version number for this model.

Returns:

An ONNX ModelProto representing the model.

writeONNXGraph

public ONNXNode writeONNXGraph(ONNXRef<?> input)

Description copied from interface: ONNXExportable

Writes this Model into OnnxMl.GraphProto.Builder inside the input's ONNXContext.

Specified by:

writeONNXGraph in interface ONNXExportable

Parameters:

input - The input to the model graph.

Returns:

the output node of the model graph.