Package com.tibco.patterns.learn.api.project

Class ModelTrainConfig

java.lang.Object

com.tibco.patterns.learn.api.project.ModelTrainConfig

public final class ModelTrainConfig
extends java.lang.Object

Defines configuration for creating and training a Learn model. Reads/writes data from/to an XML file defined by ModelTrainConfig.xsd schema. Constructs and returns ModelConfig and ConvergenceObserver objects on demand.

Constructor Summary

Constructors

Constructor	Description
ModelTrainConfig(javax.xml.validation.Schema schema, java.nio.file.Path xmlFile)	Loads the configuration from xmlFile.

Method Summary

Modifier and Type	Method	Description
COErrorRate	createCO(Partition partition, COErrorRate trainCO)	Creates convergence observer for the given partition.
static java.util.Map<Partition,ConvergenceObserver>	createDftObservers()	Creates default observers for both datasets (COErrorRate and COErrorRateMin).
ModelConfig	createModelConfig()	Reads ModelConfig from file.
java.util.Map<Partition,ConvergenceObserver>	createObservers()	Creates observers for both datasets.
boolean	isRandomOrder()
void	setModelConfig(ModelConfig mConfig)	Copies settings from mConfig to the XML structure (modelConfig node).
void	setTrainConfig(java.lang.Boolean randomOrder, java.util.Map<Partition,ConvergenceObserver> observers)	Copies settings from observers to the XML structure (trainConfig node).
java.lang.String	toString()

Methods inherited from class java.lang.Object

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

Constructor Detail

ModelTrainConfig

public ModelTrainConfig(javax.xml.validation.Schema schema,
                        java.nio.file.Path xmlFile)
                 throws javax.xml.bind.JAXBException

Loads the configuration from xmlFile. Validates that the file conforms to the schema.

Throws:

javax.xml.bind.JAXBException - if XML file validation fails.

java.lang.IllegalArgumentException - if the XML file is not found.

Method Detail

toString

public java.lang.String toString()

Overrides:

toString in class java.lang.Object

setModelConfig

public void setModelConfig(ModelConfig mConfig)

Copies settings from mConfig to the XML structure (modelConfig node). Ignores custom insert scores and remove limits for Dynamic mode.

Parameters:

mConfig - - values to initialize the model configuration in ModelTrainConfig. If null, modelConfig XML node is removed.

createModelConfig

public ModelConfig createModelConfig()

Reads ModelConfig from file. If options are not specified in the file, they will be initialized to defaults (not based on number of features).

Returns:

ModelConfig read from file, or default ModelCofig object if the file does not contain the ModelConfig section.

setTrainConfig

public void setTrainConfig(java.lang.Boolean randomOrder,
                           java.util.Map<Partition,ConvergenceObserver> observers)

Copies settings from observers to the XML structure (trainConfig node). Parameters in training observer are not copied (they are not used to stop the training).

Parameters:

randomOrder - - value for random order of training examples. If null, randomOrder XML node is removed.

observers - - values to initialize the model configuration in ModelTrainConfig. If null, trainConfig XML node is removed.

Throws:

java.lang.IllegalArgumentException - if observers contain objects other than COErrorRate or COErrorRateMin.

isRandomOrder

public boolean isRandomOrder()

Returns:

true if the order of training examples must be randomized before each training iteration or the parameter is not specified; false if it should not be randomized.

createCO

public COErrorRate createCO(Partition partition,
                            COErrorRate trainCO)

Creates convergence observer for the given partition.

Parameters:

trainCO - - observer used to create validation observer (not null). Not used to create the training observer.

Returns:

the observer to be used during training, or the default observer if it is not specified in the file.

createObservers

public java.util.Map<Partition,ConvergenceObserver> createObservers()

Creates observers for both datasets.

Returns:

convergence observers to be used for training and validation.

createDftObservers

public static java.util.Map<Partition,ConvergenceObserver> createDftObservers()

Creates default observers for both datasets (COErrorRate and COErrorRateMin).

Returns:

convergence observers to be used for training and validation.