com.pervasive.datarush.analytics.regression

Class SumOfSquares

java.lang.Object

com.pervasive.datarush.operators.AbstractLogicalOperator

com.pervasive.datarush.operators.StreamingOperator

com.pervasive.datarush.operators.ExecutableOperator

com.pervasive.datarush.analytics.regression.SumOfSquares

All Implemented Interfaces:

LogicalOperator

public final class SumOfSquares
extends ExecutableOperator

Compute the sum of squares for the given fields of the input data. The inner products are calculated distributed with a reduction at the end to produce the sum of squares matrix.

Note that the fields must be of type double or be assignable to a double type.

Constructor Summary

Constructors

Constructor and Description
SumOfSquares() Default constructor.
SumOfSquares(List<String> fieldNames) Construct with the given field names.

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) Compute the inner product for each distributed dataflow.
List<String>	getFieldNames() Get the list of field names to apply sum of squares.
RecordPort	getInput() Get the input port for this operator.
SimpleModelPort<org.apache.commons.math.linear.Array2DRowRealMatrix>	getOutput() Get the output model port of this operator.
void	setFieldNames(List<String> fieldNames) Set the list of fields to apply sum of squares.

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

Constructor Detail

SumOfSquares

public SumOfSquares()

Default constructor. The list of fields to use must be specified with setFieldNames(List).

SumOfSquares

public SumOfSquares(List<String> fieldNames)

Construct with the given field names.

Parameters:

fieldNames - names of fields compute sum of squares

Method Detail

getInput

public RecordPort getInput()

Get the input port for this operator. This port contains the data used in the sum of squares calculation.

Returns:

input port

getOutput

public SimpleModelPort<org.apache.commons.math.linear.Array2DRowRealMatrix> getOutput()

Get the output model port of this operator. This port will contain the sum of squares for the specified fields in a matrix. Only a single token will be available on this output port.

Returns:

output model port

getFieldNames

public List<String> getFieldNames()

Get the list of field names to apply sum of squares.

Returns:

field names

setFieldNames

public void setFieldNames(List<String> fieldNames)

Set the list of fields to apply sum of squares. The field names must be valid names within the schema of the input port. The fields types must be compatible with the double type.

Parameters:

fieldNames - field names

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)

Compute the inner product for each distributed dataflow. These products are merged together in a reduction step to produce the sum of squares for the input data. The merge takes place in the merge handler for the matrix output port.

Specified by:

execute in class ExecutableOperator

Parameters:

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