jsat.datatransform

Class JLTransform

java.lang.Object

jsat.datatransform.DataTransformBase

jsat.datatransform.JLTransform

All Implemented Interfaces:

java.io.Serializable, java.lang.Cloneable, DataTransform, Parameterized

public class JLTransform
extends DataTransformBase

The Johnson-Lindenstrauss (JL) Transform is a type of random projection down to a lower dimensional space. The goal is, with a high probability, to keep the Euclidean distances between points approximately the same in the original and projected space.
The JL lemma, with a high probability, bounds the error of a distance computation between two points u and v in the lower dimensional space by (1 ± ε) d(u, v)², where d is the Euclidean distance. It works best for very high dimension problems, 1000 or more.
For more information see:
Achlioptas, D. (2003). Database-friendly random projections: Johnson-Lindenstrauss with binary coins. Journal of Computer and System Sciences, 66(4), 671–687. doi:10.1016/S0022-0000(03)00025-4

See Also:

Serialized Form

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	JLTransform.TransformMode Determines which distribution to construct the transform matrix from

Constructor Summary

Constructors

Constructor and Description
JLTransform() Creates a new JL Transform that uses a target dimension of 50 features.
JLTransform(int k) Creates a new JL Transform
JLTransform(int k, JLTransform.TransformMode mode) Creates a new JL Transform
JLTransform(int k, JLTransform.TransformMode mode, boolean inMemory) Creates a new JL Transform

Method Summary

Modifier and Type	Method and Description
DataTransform	clone()
void	fit(DataSet data) Fits this transform to the given dataset.
JLTransform.TransformMode	getMode()
int	getProjectedDimension()
static Distribution	guessProjectedDimension(DataSet d)
boolean	isInMemory()
void	setInMemory(boolean inMemory) Sets whether or not the transform matrix is stored explicitly in memory or not.
void	setMode(JLTransform.TransformMode mode) The JL transform uses a random matrix to project the data, and the mode controls which method is used to construct this matrix.
void	setProjectedDimension(int k) Sets the target dimension size to use for the output
DataPoint	transform(DataPoint dp) Returns a new data point that is a transformation of the original data point.

Methods inherited from class java.lang.Object

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

Methods inherited from interface jsat.parameters.Parameterized

getParameter, getParameters

Constructor Detail

JLTransform

public JLTransform()

Creates a new JL Transform that uses a target dimension of 50 features. This may not be optimal for any particular dataset.

Parameters:

k - the target dimension size

JLTransform

public JLTransform(int k)

Creates a new JL Transform

Parameters:

k - the target dimension size

JLTransform

public JLTransform(int k,
                   JLTransform.TransformMode mode)

Creates a new JL Transform

Parameters:

k - the target dimension size

mode - how to construct the transform

rand - the source of randomness

JLTransform

public JLTransform(int k,
                   JLTransform.TransformMode mode,
                   boolean inMemory)

Creates a new JL Transform

Parameters:

k - the target dimension size

mode - how to construct the transform

inMemory - if false, the matrix will be stored in O(1) memory at the cost of execution time.

Method Detail

fit

public void fit(DataSet data)

Description copied from interface: DataTransform

Fits this transform to the given dataset. Some transforms can only be learned from classification or regression datasets. If an incompatible dataset type is given, a FailedToFitException exception may be thrown.

Parameters:

data - the dataset to fir this transform to

setMode

public void setMode(JLTransform.TransformMode mode)

The JL transform uses a random matrix to project the data, and the mode controls which method is used to construct this matrix.

Parameters:

mode - how to construct the transform

getMode

public JLTransform.TransformMode getMode()

Returns:

how to construct the transform

setInMemory

public void setInMemory(boolean inMemory)

Sets whether or not the transform matrix is stored explicitly in memory or not. Explicit storage is often faster, but can be prohibitive for large datasets

Parameters:

inMemory - true to explicitly store the transform matrix, false to re-create it on the fly as needed

isInMemory

public boolean isInMemory()

Returns:

true if this object will explicitly store the transform matrix, false to re-create it on the fly as needed

setProjectedDimension

public void setProjectedDimension(int k)

Sets the target dimension size to use for the output

Parameters:

k - the dimension after apply the transform

getProjectedDimension

public int getProjectedDimension()

Returns:

the dimension after apply the transform

guessProjectedDimension

public static Distribution guessProjectedDimension(DataSet d)

transform

public DataPoint transform(DataPoint dp)

Description copied from interface: DataTransform

Returns a new data point that is a transformation of the original data point. This new data point is a different object, but may contain the same references as the original data point. It is not guaranteed that you can mutate the transformed point without having a side effect on the original point.

Parameters:

dp - the data point to apply a transformation to

Returns:

a transformed data point

clone

public DataTransform clone()

Specified by:

clone in interface DataTransform

Specified by:

clone in class DataTransformBase