org.neuroph.core.data

Class DataSet

java.lang.Object

org.neuroph.core.data.DataSet

All Implemented Interfaces:

Serializable, Iterable<DataSetRow>, javax.visrec.ml.data.DataSet<DataSetRow>

Direct Known Subclasses:

BufferedDataSet

public class DataSet
extends Object
implements javax.visrec.ml.data.DataSet<DataSetRow>, Serializable

This class represents a collection of data rows (DataSetRow instances) used for training and testing neural network. TODO: add logging

Author:

Zoran Sevarac

See Also:

http://openforecast.sourceforge.net/docs/net/sourceforge/openforecast/DataSet.html, Serialized Form

Nested Class Summary

Nested classes/interfaces inherited from interface javax.visrec.ml.data.DataSet

javax.visrec.ml.data.DataSet.Column, javax.visrec.ml.data.DataSet.ColumnType

Constructor Summary

Constructors

Constructor and Description
DataSet(int inputSize) Creates an instance of new empty training set
DataSet(int inputSize, int outputSize) Creates an instance of new empty training set

Method Summary

Modifier and Type	Method and Description
DataSet	add(DataSetRow row) Adds new row row to this data set
void	add(double[] input, double[] output) Adds a new dataset row with specified input and output
void	add(int index, DataSetRow row)
void	addRow(double[] input) Adds a new dataset row with specified input
void	clear() Removes all alements from training set
static DataSet	createFromFile(String filePath, int inputsCount, int outputsCount, String delimiter) Creates and returns data set from specified csv file
static DataSet	createFromFile(String filePath, int inputsCount, int outputsCount, String delimiter, boolean loadColumnNames) Creates and returns data set from specified csv file
DataSet[]	createTrainingAndTestSubsets(double trainSetPercent, double testSetPercent) Returns training and test subsets in the specified percent ratio
DataSetRow	get(int index)
String	getColumnName(int idx)
String[]	getColumnNames()
DataSetColumnType	getColumnType(int index)
DataSetColumnType[]	getColumnTypes()
String	getFilePath() Returns full file path for this training set
int	getInputSize() Returns input vector size of training elements in this training set This method is implementation of EngineIndexableSet interface, and it is added to provide compatibility with Encog data sets and FlatNetwork
List<DataSetRow>	getItems()
String	getLabel() Returns label for this training set
int	getOutputSize() Returns output vector size of training elements in this training set.
DataSetRow	getRowAt(int idx) Returns training row at specified index position
List<DataSetRow>	getRows() Returns elements of this training set
int	indexOf(Object row)
boolean	isEmpty() Returns true if training set is empty, false otherwise
boolean	isSupervised() Returns true if data set is supervised, false otherwise
Iterator<DataSetRow>	iterator() Returns Iterator for iterating training elements collection
static DataSet	load(String filePath) Loads training set from the specified file TODO: throw checked exceptionse here
DataSetRow	remove(int index)
boolean	remove(Object row)
void	removeRowAt(int idx) Removes training row at specified index position
DataSet[]	sample(Sampling sampling)
void	save() Saves this training set to file specified in its filePath field
void	save(String filePath) Saves this training set to the specified file
void	saveAsTxt(String filePath, String delimiter)
void	setColumnName(int idx, String columnName)
void	setColumnNames(String[] columnNames)
void	setColumnType(int index, DataSetColumnType columnType) Sets column type for the given index.
void	setFilePath(String filePath) Sets full file path for this training set
void	setLabel(String label) Sets label for this training set
void	shuffle()
int	size() Returns number of training elements in this training set set
DataSet[]	split(double... parts) Splits data sets into parts of specified sizes.
DataSet[]	split(int numParts) Splits data set into specified number of parts and returns them as a list.
javax.visrec.ml.data.DataSet<DataSetRow>[]	split(int numParts, Random rnd)
javax.visrec.ml.data.DataSet<DataSetRow>[]	split(Random rnd, double... parts)
String	toCSV() Returns enire dataset in csv format
String	toString() Returns string representation of this data set

Methods inherited from class java.lang.Object

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

Methods inherited from interface javax.visrec.ml.data.DataSet

addAll, shuffle, split

Methods inherited from interface java.lang.Iterable

forEach, spliterator

Constructor Detail

DataSet

public DataSet(int inputSize)

Creates an instance of new empty training set

Parameters:

inputSize -

DataSet

public DataSet(int inputSize,
               int outputSize)

Creates an instance of new empty training set

Parameters:

inputSize - Length of the input vector

outputSize - Length of the output vector

Method Detail

add

public DataSet add(DataSetRow row)
            throws VectorSizeMismatchException

Adds new row row to this data set

Specified by:

add in interface javax.visrec.ml.data.DataSet<DataSetRow>

Parameters:

row - data set row to add

Returns:

Throws:

VectorSizeMismatchException

addRow

public void addRow(double[] input)

Adds a new dataset row with specified input

Parameters:

input -

add

public void add(double[] input,
                double[] output)

Adds a new dataset row with specified input and output

Parameters:

input -

output -

removeRowAt

public void removeRowAt(int idx)

Removes training row at specified index position

Parameters:

idx - position of row to remove

iterator

public Iterator<DataSetRow> iterator()

Returns Iterator for iterating training elements collection

Specified by:

iterator in interface Iterable<DataSetRow>

Returns:

Iterator for iterating training elements collection

getRows

public List<DataSetRow> getRows()

Returns elements of this training set

Returns:

training elements

getRowAt

public DataSetRow getRowAt(int idx)

Returns training row at specified index position

Parameters:

idx - index position of training row to return

Returns:

training row at specified index position

clear

public void clear()

Removes all alements from training set

Specified by:

clear in interface javax.visrec.ml.data.DataSet<DataSetRow>

isEmpty

public boolean isEmpty()

Returns true if training set is empty, false otherwise

Specified by:

isEmpty in interface javax.visrec.ml.data.DataSet<DataSetRow>

Returns:

true if training set is empty, false otherwise

isSupervised

public boolean isSupervised()

Returns true if data set is supervised, false otherwise

Returns:

size

public int size()

Returns number of training elements in this training set set

Specified by:

size in interface javax.visrec.ml.data.DataSet<DataSetRow>

Returns:

number of training elements in this training set set

getLabel

public String getLabel()

Returns label for this training set

Returns:

label for this training set

setLabel

public void setLabel(String label)

Sets label for this training set

Parameters:

label - label for this training set

getColumnNames

public String[] getColumnNames()

setColumnNames

public void setColumnNames(String[] columnNames)

getColumnName

public String getColumnName(int idx)

setColumnName

public void setColumnName(int idx,
                          String columnName)

getColumnTypes

public DataSetColumnType[] getColumnTypes()

getColumnType

public DataSetColumnType getColumnType(int index)

setColumnType

public void setColumnType(int index,
                          DataSetColumnType columnType)

Sets column type for the given index.

Parameters:

index - Index of the column in the row.

columnType - Column type to set, nominal or numeric.

setFilePath

public void setFilePath(String filePath)

Sets full file path for this training set

Parameters:

filePath -

getFilePath

public String getFilePath()

Returns full file path for this training set

Returns:

full file path for this training set

toString

public String toString()

Returns string representation of this data set

Overrides:

toString in class Object

Returns:

string representation of this data set

toCSV

public String toCSV()

Returns enire dataset in csv format

Returns:

save

public void save(String filePath)

Saves this training set to the specified file

Parameters:

filePath -

save

public void save()

Saves this training set to file specified in its filePath field

saveAsTxt

public void saveAsTxt(String filePath,
                      String delimiter)

load

public static DataSet load(String filePath)

Loads training set from the specified file TODO: throw checked exceptionse here

Parameters:

filePath - training set file

Returns:

loded training set

createFromFile

public static DataSet createFromFile(String filePath,
                                     int inputsCount,
                                     int outputsCount,
                                     String delimiter,
                                     boolean loadColumnNames)

Creates and returns data set from specified csv file

Parameters:

filePath - path to csv dataset file to import

inputsCount - number of inputs

outputsCount - number of outputs

delimiter - delimiter of values

loadColumnNames - true if csv file contains column names in first line, false otherwise

Returns:

instance of dataset with values from specified file TODO: try with resources, provide information on exact line of error if format is not good in NumberFormatException

createFromFile

public static DataSet createFromFile(String filePath,
                                     int inputsCount,
                                     int outputsCount,
                                     String delimiter)

Creates and returns data set from specified csv file

Parameters:

filePath - path to csv dataset file to import

inputsCount - number of inputs

outputsCount - number of outputs

delimiter - delimiter of values

Returns:

instance of dataset with values from specified file

createTrainingAndTestSubsets

public DataSet[] createTrainingAndTestSubsets(double trainSetPercent,
                                              double testSetPercent)

Returns training and test subsets in the specified percent ratio

Parameters:

trainSetPercent -

testSetPercent -

Returns:

split

public DataSet[] split(int numParts)

Splits data set into specified number of parts and returns them as a list.

Specified by:

split in interface javax.visrec.ml.data.DataSet<DataSetRow>

Parameters:

numParts -

Returns:

split

public DataSet[] split(double... parts)

Splits data sets into parts of specified sizes. Sum of parts must be equal one, or exception will be thrown.

Specified by:

split in interface javax.visrec.ml.data.DataSet<DataSetRow>

Parameters:

parts -

Returns:

sample

public DataSet[] sample(Sampling sampling)

getOutputSize

public int getOutputSize()

Returns output vector size of training elements in this training set.

getInputSize

public int getInputSize()

Returns input vector size of training elements in this training set This method is implementation of EngineIndexableSet interface, and it is added to provide compatibility with Encog data sets and FlatNetwork

Returns:

shuffle

public void shuffle()

Specified by:

shuffle in interface javax.visrec.ml.data.DataSet<DataSetRow>

remove

public boolean remove(Object row)

get

public DataSetRow get(int index)

Specified by:

get in interface javax.visrec.ml.data.DataSet<DataSetRow>

add

public void add(int index,
                DataSetRow row)

remove

public DataSetRow remove(int index)

indexOf

public int indexOf(Object row)

getItems

public List<DataSetRow> getItems()

Specified by:

getItems in interface javax.visrec.ml.data.DataSet<DataSetRow>

split

public javax.visrec.ml.data.DataSet<DataSetRow>[] split(int numParts,
                                                        Random rnd)

Specified by:

split in interface javax.visrec.ml.data.DataSet<DataSetRow>

split

public javax.visrec.ml.data.DataSet<DataSetRow>[] split(Random rnd,
                                                        double... parts)

Specified by:

split in interface javax.visrec.ml.data.DataSet<DataSetRow>