jsat.datatransform.visualization

Class LargeViz

java.lang.Object

jsat.datatransform.visualization.LargeViz

All Implemented Interfaces:

java.io.Serializable, java.lang.Cloneable, VisualizationTransform

public class LargeViz
extends java.lang.Object
implements VisualizationTransform

LargeViz is an algorithm for creating low dimensional embeddings for visualization. It is meant to be faster and better quality than t-SNE without requiring any parameter tuning to get good results. LargeViz is related to t-SNE in how the neighbor graph is constructed, and the perplexity parameter for LargeViz has the same meaning and impact as the perplexity parameter in t-SNE.

NOTE: The origina LargeViz paper includes a faster scheme for approximately constructing the nearest neighbor graph. This is not yet implemented, but has no negative impact on the quality of the result.
See: Tang, J., Liu, J., Zhang, M., & Mei, Q. (2016). Visualizing Large-scale and High-dimensional Data. In Proceedings of the 25th International Conference on World Wide Web (pp. 287–297). Republic and Canton of Geneva, Switzerland: International World Wide Web Conferences Steering Committee. doi:10.1145/2872427.2883041

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
LargeViz()

Method Summary

Modifier and Type	Method and Description
double	getGamma()
int	getNegativeSamples()
double	getPerplexity()
int	getTargetDimension()
void	setDistanceMetricEmbedding(DistanceMetric dm) Sets the distance metric to use for the embedded space.
void	setDistanceMetricSource(DistanceMetric dm) Sets the distance metric to use for the original space.
void	setGamma(double gamma) Gamma controls the negative weight assigned to negative edges in the optimization problem.
void	setNegativeSamples(int M) Sets the number of negative neighbor samples to obtain for each data point.
void	setPerplexity(double perplexity) Sets the target perplexity of the gaussian used over each data point.
boolean	setTargetDimension(int target) Sets the target dimension to embed new dataset to.
<Type extends DataSet> Type	transform(DataSet<Type> d, boolean parallel) Transforms the given data set, returning a dataset of the same type.

Methods inherited from class java.lang.Object

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

Methods inherited from interface jsat.datatransform.visualization.VisualizationTransform

transform

Constructor Detail

LargeViz

public LargeViz()

Method Detail

setPerplexity

public void setPerplexity(double perplexity)

Sets the target perplexity of the gaussian used over each data point. The perplexity can be thought of as a quasi desired number of nearest neighbors to be considered, but is adapted based on the distribution of the data. Increasing the perplexity can increase the amount of time it takes to get an embedding. Using a value in the range of [5, 100] is recommended.

Parameters:

perplexity - the quasi number of neighbors to consider for each data point

getPerplexity

public double getPerplexity()

Returns:

the target perplexity to use for each data point

setDistanceMetricSource

public void setDistanceMetricSource(DistanceMetric dm)

Sets the distance metric to use for the original space. This will determine the target nearest neighbors to keep close to each other in the embedding space

Parameters:

dm - the distance metric to use

setDistanceMetricEmbedding

public void setDistanceMetricEmbedding(DistanceMetric dm)

Sets the distance metric to use for the embedded space. This will determine the actual nearest neighbors as the occur in the embedded space.

Parameters:

dm - the distance metric to use

setNegativeSamples

public void setNegativeSamples(int M)

Sets the number of negative neighbor samples to obtain for each data point. The default recommended value is 5.

Parameters:

M - the number of negative samples to use for each update

getNegativeSamples

public int getNegativeSamples()

Returns:

the number of negative samples to use for each update

setGamma

public void setGamma(double gamma)

Gamma controls the negative weight assigned to negative edges in the optimization problem. Large values will place a higher emphasis on separating non-neighbors in the embedded space. The default recommend value is 7.

Parameters:

gamma - the weight for negative edge samples

getGamma

public double getGamma()

Returns:

the weight for negative edge samples

getTargetDimension

public int getTargetDimension()

Specified by:

getTargetDimension in interface VisualizationTransform

Returns:

the number of dimensions that a dataset will be embedded down to

setTargetDimension

public boolean setTargetDimension(int target)

Description copied from interface: VisualizationTransform

Sets the target dimension to embed new dataset to. Many visualization methods may only support a target of 2 or 3 dimensions, or only one of those options. For that reason a boolean value will be returned indicating if the target size was acceptable. If not, no change to the object will occur.

Specified by:

setTargetDimension in interface VisualizationTransform

Parameters:

target - the new target dimension size when VisualizationTransform.transform(jsat.DataSet) is called.

Returns:

true if this transform supports that dimension and it was set, false if the target dimension is unsupported and the previous value will be used instead.

transform

public <Type extends DataSet> Type transform(DataSet<Type> d,
                                             boolean parallel)

Description copied from interface: VisualizationTransform

Transforms the given data set, returning a dataset of the same type.

Specified by:

transform in interface VisualizationTransform

Type Parameters:

Type - the dataset type

Parameters:

d - the data set to transform

parallel - true if transform should be done in parallel, or false if it should use a single thread.

Returns:

the lower dimension dataset for visualization.