org.deeplearning4j.datasets.datavec

Class RecordReaderMultiDataSetIterator.Builder

java.lang.Object

org.deeplearning4j.datasets.datavec.RecordReaderMultiDataSetIterator.Builder

Enclosing class:

RecordReaderMultiDataSetIterator

public static class RecordReaderMultiDataSetIterator.Builder
extends Object

Constructor Summary

Constructors

Constructor and Description
Builder(int batchSize)

Method Summary

Modifier and Type	Method and Description
RecordReaderMultiDataSetIterator.Builder	addInput(String readerName) Set as an input, the entire contents (all columns) of the RecordReader or SequenceRecordReader
RecordReaderMultiDataSetIterator.Builder	addInput(String readerName, int columnFirst, int columnLast) Set as an input, a subset of the specified RecordReader or SequenceRecordReader
RecordReaderMultiDataSetIterator.Builder	addInputOneHot(String readerName, int column, int numClasses) Add as an input a single column from the specified RecordReader / SequenceRecordReader The assumption is that the specified column contains integer values in range 0..numClasses-1; this integer will be converted to a one-hot representation
RecordReaderMultiDataSetIterator.Builder	addOutput(String readerName) Set as an output, the entire contents (all columns) of the RecordReader or SequenceRecordReader
RecordReaderMultiDataSetIterator.Builder	addOutput(String readerName, int columnFirst, int columnLast) Add an output, with a subset of the columns from the named RecordReader or SequenceRecordReader
RecordReaderMultiDataSetIterator.Builder	addOutputOneHot(String readerName, int column, int numClasses) An an output, where the output is taken from a single column from the specified RecordReader / SequenceRecordReader The assumption is that the specified column contains integer values in range 0..numClasses-1; this integer will be converted to a one-hot representation (usually for classification)
RecordReaderMultiDataSetIterator.Builder	addReader(String readerName, RecordReader recordReader) Add a RecordReader for use in .addInput(...) or .addOutput(...)
RecordReaderMultiDataSetIterator.Builder	addSequenceReader(String seqReaderName, SequenceRecordReader seqRecordReader) Add a SequenceRecordReader for use in .addInput(...) or .addOutput(...)
RecordReaderMultiDataSetIterator	build() Create the RecordReaderMultiDataSetIterator
RecordReaderMultiDataSetIterator.Builder	sequenceAlignmentMode(RecordReaderMultiDataSetIterator.AlignmentMode alignmentMode) Set the sequence alignment mode for all sequences
RecordReaderMultiDataSetIterator.Builder	timeSeriesRandomOffset(boolean timeSeriesRandomOffset, long rngSeed) For use with timeseries trained with tbptt In a given minbatch, shorter time series are padded and appropriately masked to be the same length as the longest time series.

Methods inherited from class java.lang.Object

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

Constructor Detail

Builder

public Builder(int batchSize)

Parameters:

batchSize - The batch size for the RecordReaderMultiDataSetIterator

Method Detail

addReader

public RecordReaderMultiDataSetIterator.Builder addReader(String readerName,
                                                          RecordReader recordReader)

Add a RecordReader for use in .addInput(...) or .addOutput(...)

Parameters:

readerName - Name of the reader (for later reference)

recordReader - RecordReader

addSequenceReader

public RecordReaderMultiDataSetIterator.Builder addSequenceReader(String seqReaderName,
                                                                  SequenceRecordReader seqRecordReader)

Add a SequenceRecordReader for use in .addInput(...) or .addOutput(...)

Parameters:

seqReaderName - Name of the sequence reader (for later reference)

seqRecordReader - SequenceRecordReader

sequenceAlignmentMode

public RecordReaderMultiDataSetIterator.Builder sequenceAlignmentMode(RecordReaderMultiDataSetIterator.AlignmentMode alignmentMode)

Set the sequence alignment mode for all sequences

addInput

public RecordReaderMultiDataSetIterator.Builder addInput(String readerName)

Set as an input, the entire contents (all columns) of the RecordReader or SequenceRecordReader

addInput

public RecordReaderMultiDataSetIterator.Builder addInput(String readerName,
                                                         int columnFirst,
                                                         int columnLast)

Set as an input, a subset of the specified RecordReader or SequenceRecordReader

Parameters:

readerName - Name of the reader

columnFirst - First column index, inclusive

columnLast - Last column index, inclusive

addInputOneHot

public RecordReaderMultiDataSetIterator.Builder addInputOneHot(String readerName,
                                                               int column,
                                                               int numClasses)

Add as an input a single column from the specified RecordReader / SequenceRecordReader The assumption is that the specified column contains integer values in range 0..numClasses-1; this integer will be converted to a one-hot representation

Parameters:

readerName - Name of the RecordReader or SequenceRecordReader

column - Column that contains the index

numClasses - Total number of classes

addOutput

public RecordReaderMultiDataSetIterator.Builder addOutput(String readerName)

Set as an output, the entire contents (all columns) of the RecordReader or SequenceRecordReader

addOutput

public RecordReaderMultiDataSetIterator.Builder addOutput(String readerName,
                                                          int columnFirst,
                                                          int columnLast)

Add an output, with a subset of the columns from the named RecordReader or SequenceRecordReader

Parameters:

readerName - Name of the reader

columnFirst - First column index

columnLast - Last column index (inclusive)

addOutputOneHot

public RecordReaderMultiDataSetIterator.Builder addOutputOneHot(String readerName,
                                                                int column,
                                                                int numClasses)

An an output, where the output is taken from a single column from the specified RecordReader / SequenceRecordReader The assumption is that the specified column contains integer values in range 0..numClasses-1; this integer will be converted to a one-hot representation (usually for classification)

Parameters:

readerName - Name of the RecordReader / SequenceRecordReader

column - index of the column

numClasses - Number of classes

timeSeriesRandomOffset

public RecordReaderMultiDataSetIterator.Builder timeSeriesRandomOffset(boolean timeSeriesRandomOffset,
                                                                       long rngSeed)

For use with timeseries trained with tbptt In a given minbatch, shorter time series are padded and appropriately masked to be the same length as the longest time series. Cases with a skewed distrbution of lengths can result in the last few updates from the time series coming from mostly masked time steps. timeSeriesRandomOffset randomly offsettsthe time series + masking appropriately to address this

Parameters:

timeSeriesRandomOffset, - "true" to randomly offset time series within a minibatch

rngSeed - seed for reproducibility

build

public RecordReaderMultiDataSetIterator build()

Create the RecordReaderMultiDataSetIterator