com.pervasive.datarush.analytics.svm.learner

Class SVMLearner

java.lang.Object

com.pervasive.datarush.operators.AbstractLogicalOperator

com.pervasive.datarush.operators.StreamingOperator

com.pervasive.datarush.operators.ExecutableOperator

com.pervasive.datarush.analytics.svm.learner.SVMLearner

All Implemented Interfaces:

LogicalOperator, RecordSinkOperator, SinkOperator<RecordPort>

public final class SVMLearner
extends ExecutableOperator
implements RecordSinkOperator

Builds a PMMLSupportVectorMachineModel from an input dataset.

NOTE: This operator is implemented as a wrapper for LIBSVM. Please refer to the LIBSVM documentation for additional information regarding the parameters.

NOTE: this operator is non-parallel

Constructor Summary

Constructors

Constructor and Description
SVMLearner() Learns an SVM model from a training dataset.

Method Summary

Methods

Modifier and Type	Method and Description
protected void	computeMetadata(StreamingMetadataContext ctx) Implementations must adhere to the following contracts
protected void	execute(ExecutionContext ctx) Executes the operator.
double	getEpsilon() Returns the tolerance for termination criteria.
List<String>	getIncludedColumns() Returns the list of columns to include for the purpose of building the model.
RecordPort	getInput() The input data.
KernelType	getKernelType() Returns the kernel and associated parameters to use.
PMMLPort	getModel() Returns a dataflow variable that will contain the PMML model.
double	getSvmCacheSizeMB() Returns the cache size.
SVMType	getType() Returns the type of SVM model to build.
static boolean	isSilent() Our default SVM library will send regular status reports to System.out.
void	setEpsilon(double epsilon) Sets the tolerance for termination criteria.
void	setIncludedColumns(List<String> includedColumns) Sets the list of columns to include.
void	setKernelType(KernelType kernelType) Sets the kernel and associated parameters to use.
static void	setSilent(boolean quiet) Our default SVM library will send regular status reports to System.out.
void	setSvmCacheSizeMB(double svmCacheSizeMB) Sets the cache size.
void	setType(SVMType type) Sets the type of SVM model to build.

Methods inherited from class com.pervasive.datarush.operators.ExecutableOperator

cloneForExecution, getNumInputCopies, getPortSettings, handleInactiveOutput

Methods inherited from class com.pervasive.datarush.operators.AbstractLogicalOperator

disableParallelism, getInputPorts, getOutputPorts, newInput, newInput, newOutput, newRecordInput, newRecordInput, newRecordOutput, notifyError

Methods inherited from class java.lang.Object

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

Methods inherited from interface com.pervasive.datarush.operators.LogicalOperator

disableParallelism, getInputPorts, getOutputPorts

Constructor Detail

SVMLearner

public SVMLearner()

Learns an SVM model from a training dataset.

Method Detail

getModel

public PMMLPort getModel()

Returns a dataflow variable that will contain the PMML model.

Returns:

a dataflow variable that will contain the PMML model.

getInput

public RecordPort getInput()

The input data. This contains both the independent variables and the target variable ( if applicable ). The target variable only applies to SVM's of type SVMPredictorType.

Specified by:

getInput in interface RecordSinkOperator

Specified by:

getInput in interface SinkOperator<RecordPort>

Returns:

input data port

setSilent

public static void setSilent(boolean quiet)

Our default SVM library will send regular status reports to System.out. However, it has a static property that can suppress console output. If console output is desired, setSilent(false) This method has side effects in the static variables of libsvm.svm

Parameters:

quiet - true if this SVMLearner will suppress libSVM's console output; false if it shouldn't.

isSilent

public static boolean isSilent()

Our default SVM library will send regular status reports to System.out. However, it has a static property that can suppress console output. By default, we suppress it, but it may be switched on at runtime.

Returns:

whether the SVM trainer's console output is suppressed.

getIncludedColumns

public List<String> getIncludedColumns()

Returns the list of columns to include for the purpose of building the model. An empty list means all columns of type double.

Returns:

The list of columns to include

setIncludedColumns

public void setIncludedColumns(List<String> includedColumns)

Sets the list of columns to include. An empty list means all columns of type double.

Parameters:

includedColumns - The list of columns to include

getKernelType

public final KernelType getKernelType()

Returns the kernel and associated parameters to use.

Returns:

the kernel and associated parameters to use.

setKernelType

public final void setKernelType(KernelType kernelType)

Sets the kernel and associated parameters to use.

Parameters:

kernelType - the kernel and associated parameters to use.

getEpsilon

public double getEpsilon()

Returns the tolerance for termination criteria. Defaults to 0.001. Larger values will terminate early but provide less precise results. Directly maps to the LIBSVM "-e" command line flag.

Returns:

the tolerance for termination criteria

setEpsilon

public void setEpsilon(double epsilon)

Sets the tolerance for termination criteria. Defaults to 0.001. Larger values will terminate early but provide less precise results. Directly maps to the LIBSVM "-e" command line flag.

Parameters:

epsilon - the tolerance for termination criteria

getType

public SVMType getType()

Returns the type of SVM model to build. Defaults to SVMTypeOneClass.

Returns:

the type of SVM model to build.

setType

public void setType(SVMType type)

Sets the type of SVM model to build. Defaults to SVMTypeOneClass.

Parameters:

type - the type of SVM model to build.

getSvmCacheSizeMB

public double getSvmCacheSizeMB()

Returns the cache size. Directly maps to the LIBSVM "-m" command line flag.

Returns:

the cache size.

setSvmCacheSizeMB

public void setSvmCacheSizeMB(double svmCacheSizeMB)

Sets the cache size. Directly maps to the LIBSVM "-m" command line flag.

Parameters:

svmCacheSizeMB - the cache size.

computeMetadata

protected void computeMetadata(StreamingMetadataContext ctx)

Description copied from class: StreamingOperator

Implementations must adhere to the following contracts

General

Regardless of input ports/output port types, all implementations must do the following:

1. Validation. Validation of configuration should always be performed first.
2. Declare parallelizability.. Implementations must declare parallelizability by calling StreamingMetadataContext.parallelize(ParallelismStrategy).

Input record ports

Implementations with input record ports must declare the following:

1. Required data ordering:
Implementations that have data ordering requirements must declare them by calling RecordPort.setRequiredDataOrdering(com.pervasive.datarush.operators.MetadataCalculationContext, com.pervasive.datarush.ports.record.DataOrdering), otherwise data may arrive in any order.
2. Required data distribution (only applies to parallelizable operators):
Implementations that have data distribution requirements must declare them by calling RecordPort.setRequiredDataDistribution(com.pervasive.datarush.operators.MetadataCalculationContext, com.pervasive.datarush.ports.record.DataDistribution), otherwise data will arrive in an unspecified partial distribution.

Note that if the upstream operator's output distribution/ordering is compatible with those required, we avoid a re-sort/re-distribution which is generally a very large savings from a performance standpoint. In addition, some operators may chose to query the upstream output distribution/ordering by calling RecordPort.getSourceDataDistribution(com.pervasive.datarush.operators.MetadataContext) and RecordPort.getSourceDataOrdering(com.pervasive.datarush.operators.MetadataContext). These should be viewed as a hints to help chose a more efficient algorithm. In such cases, though, operators must still declare data ordering and data distribution requirements; otherwise there is no guarantee that data will arrive sorted/distributed as required.

Output record ports

Implementations with output record ports must declare the following:

1. Type: Implementations must declare their output type by calling RecordPort.setType(com.pervasive.datarush.operators.MetadataCalculationContext, com.pervasive.datarush.types.RecordTokenType).

Implementations with output record ports may declare the following:

1. Output data ordering: Implementations that can make guarantees as to their output ordering may do so by calling RecordPort.setOutputDataOrdering(com.pervasive.datarush.operators.MetadataCalculationContext, com.pervasive.datarush.ports.record.DataOrdering)
2. Output data distribution (only applies to parallelizable operators): Implementations that can make guarantees as to their output distribution may do so by calling RecordPort.setOutputDataDistribution(com.pervasive.datarush.operators.MetadataCalculationContext, com.pervasive.datarush.ports.record.DataDistribution)

Note that both of these properties are optional; if unspecified, performance may suffer since the framework may unnecessarily re-sort/re-distributed the data.

Input model ports

In general, there is nothing special to declare for input model ports. Models are implicitly duplicated to all partitions when going from non-parallel to parallel operators. The case of a model going from a parallel to a non-parallel node is a special case of a "model reducer" operator. In the case of a model reducer, the downstream operator, must declare the following:

1. Merge handler: Model reducers must declare a merge handler by calling AbstractModelPort.setMergeHandler(com.pervasive.datarush.operators.MetadataCalculationContext, com.pervasive.datarush.ports.model.ModelMergeHandler<T>).

Note that MergeModel is a convenient, re-usable model reducer, parameterized with a merge-handler.

Output model ports

SimpleModelPort's have no associated metadata and therefore there is never any output metadata to declare. PMMLPort's, on the other hand, do have associated metadata. For all PMMLPorts, implementations must declare the following:

1. pmmlModelSpec: Implementations must declare the PMML model spec by calling PMMLPort.setPMMLModelSpec.

Specified by:

computeMetadata in class StreamingOperator

Parameters:

ctx - the context

execute

protected void execute(ExecutionContext ctx)

Description copied from class: ExecutableOperator

Executes the operator. Implementations should adhere to the following contracts:

1. Following execution, all input ports must be at end-of-data.
2. Following execution, all output ports must be at end-of-data.

Specified by:

execute in class ExecutableOperator

Parameters:

ctx - context in which to lookup physical ports bound to logical ports