org.tribuo

Interface OutputFactory<T extends Output<T>>

All Superinterfaces:

com.oracle.labs.mlrg.olcut.config.Configurable, com.oracle.labs.mlrg.olcut.provenance.Provenancable<OutputFactoryProvenance>, Serializable

All Known Implementing Classes:

AnomalyFactory, ClusteringFactory, LabelFactory, MultiLabelFactory, RegressionFactory

public interface OutputFactory<T extends Output<T>>
extends com.oracle.labs.mlrg.olcut.config.Configurable, com.oracle.labs.mlrg.olcut.provenance.Provenancable<OutputFactoryProvenance>, Serializable

An interface associated with a specific Output, which can generate the appropriate Output subclass, and OutputInfo subclass.

Must be Configurable so it can be loaded from an olcut config file.

Method Summary

Modifier and Type	Method and Description
ImmutableOutputInfo<T>	constructInfoForExternalModel(Map<T,Integer> mapping) Creates an ImmutableOutputInfo from the supplied mapping.
MutableOutputInfo<T>	generateInfo() Generates the appropriate MutableOutputInfo so the output values can be tracked by a Dataset or other aggregate.
<V> T	generateOutput(V label) Parses the V and generates the appropriate Output value.
default <V> List<T>	generateOutputs(List<V> inputs) Generate a list of outputs from the supplied list of inputs.
Evaluator<T,? extends Evaluation<T>>	getEvaluator() Gets an Evaluator suitable for measuring performance of predictions for the Output subclass.
T	getUnknownOutput() Returns the singleton unknown output of type T which can be used for prediction time examples.
static <T extends Output<T>> void	validateMapping(Map<T,Integer> mapping) Validates that the mapping can be used as an output info, i.e.

Methods inherited from interface com.oracle.labs.mlrg.olcut.config.Configurable

postConfig

Methods inherited from interface com.oracle.labs.mlrg.olcut.provenance.Provenancable

getProvenance

Method Detail

generateOutput

<V> T generateOutput(V label)

Parses the V and generates the appropriate Output value.

Most implementations call toString on the label before parsing it, but this is not required.

Type Parameters:

V - The type of the input value.

Parameters:

label - An input value.

Returns:

The parsed Output as an instance of T.

getUnknownOutput

T getUnknownOutput()

Returns the singleton unknown output of type T which can be used for prediction time examples.

Returns:

An unknown output.

generateInfo

MutableOutputInfo<T> generateInfo()

Generates the appropriate MutableOutputInfo so the output values can be tracked by a Dataset or other aggregate.

Returns:

The appropriate subclass of MutableOutputInfo initialised to zero.

constructInfoForExternalModel

ImmutableOutputInfo<T> constructInfoForExternalModel(Map<T,Integer> mapping)

Creates an ImmutableOutputInfo from the supplied mapping. Requires that the mapping is dense in the integers [0,mapping.size()) and each mapping is unique.

This call is used to import external models, and should not be used for other purposes.

Parameters:

mapping - The mapping to use.

Returns:

The appropriate subclass of ImmutableOutputInfo with a single observation of each element.

getEvaluator

Evaluator<T,? extends Evaluation<T>> getEvaluator()

Gets an Evaluator suitable for measuring performance of predictions for the Output subclass.

Evaluator instances are thread safe and immutable, and commonly this is a singleton stored in the OutputFactory implementation.

Returns:

An evaluator.

generateOutputs

default <V> List<T> generateOutputs(List<V> inputs)

Generate a list of outputs from the supplied list of inputs.

Makes inputs.size() calls to generateOutput(V).

Type Parameters:

V - The type of the inputs

Parameters:

inputs - The list to convert.

Returns:

A list of outputs.

validateMapping

static <T extends Output<T>> void validateMapping(Map<T,Integer> mapping)

Validates that the mapping can be used as an output info, i.e. that it is dense in the region [0,mapping.size()) - meaning no duplicate ids, each id 0 through mapping.size() is used, and there are no negative ids.

Type Parameters:

T - The type of the output.

Parameters:

mapping - The mapping to use.