edu.stanford.nlp.loglinear.model

Class ConcatVectorNamespace

java.lang.Object

edu.stanford.nlp.loglinear.model.ConcatVectorNamespace

public class ConcatVectorNamespace
extends java.lang.Object

Created on 10/20/15.

Author:

keenon

This is a wrapper function to keep a namespace of namespace of recognized features, so that building a set of ConcatVectors for featurizing a model is easier and more intuitive. It's actually quite simple, and threadsafe.

Constructor Summary

Constructors

Constructor and Description
ConcatVectorNamespace()

Method Summary

Modifier and Type	Method and Description
void	debugVector(ConcatVector vector, java.io.BufferedWriter bw) This prints out a ConcatVector by mapping to the namespace, to make debugging learning algorithms easier.
int	ensureFeature(java.lang.String featureName) An optimization, this lets clients inform the ConcatVectorNamespace of how many features to expect, so that we can avoid resizing ConcatVectors.
int	ensureSparseFeature(java.lang.String featureName, java.lang.String index) An optimization, this lets clients inform the ConcatVectorNamespace of how many sparse feature components to expect, again so that we can avoid resizing ConcatVectors.
ConcatVector	newVector() Creates a new vector that is appropriately sized to accommodate all the features that have been named so far.
ConcatVector	newWeightsVector() This constructs a fresh vector that is sized correctly to accommodate all the known sparse values for vectors that are possibly sparse.
void	setDenseFeature(ConcatVector vector, java.lang.String featureName, double[] value) This adds a dense feature to a vector, setting the appropriate component of the given vector to the passed in value.
void	setSparseFeature(ConcatVector vector, java.lang.String featureName, java.lang.String index, double value) This adds a sparse feature to a vector, setting the appropriate component of the given vector to the passed in value.

Methods inherited from class java.lang.Object

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

Constructor Detail

ConcatVectorNamespace

public ConcatVectorNamespace()

Method Detail

newVector

public ConcatVector newVector()

Creates a new vector that is appropriately sized to accommodate all the features that have been named so far.

Returns:

a new, empty ConcatVector

newWeightsVector

public ConcatVector newWeightsVector()

This constructs a fresh vector that is sized correctly to accommodate all the known sparse values for vectors that are possibly sparse.

Returns:

a new, internally correctly sized ConcatVector that will work correctly as weights for features from this namespace;

ensureFeature

public int ensureFeature(java.lang.String featureName)

An optimization, this lets clients inform the ConcatVectorNamespace of how many features to expect, so that we can avoid resizing ConcatVectors.

Parameters:

featureName - the feature to add to our index

ensureSparseFeature

public int ensureSparseFeature(java.lang.String featureName,
                               java.lang.String index)

An optimization, this lets clients inform the ConcatVectorNamespace of how many sparse feature components to expect, again so that we can avoid resizing ConcatVectors.

Parameters:

featureName - the feature to use in our index

index - the sparse value to ensure is available

setDenseFeature

public void setDenseFeature(ConcatVector vector,
                            java.lang.String featureName,
                            double[] value)

This adds a dense feature to a vector, setting the appropriate component of the given vector to the passed in value.

Parameters:

vector - the vector

featureName - the feature whose value to set

value - the value we want to set this vector to

setSparseFeature

public void setSparseFeature(ConcatVector vector,
                             java.lang.String featureName,
                             java.lang.String index,
                             double value)

This adds a sparse feature to a vector, setting the appropriate component of the given vector to the passed in value.

Parameters:

vector - the vector

featureName - the feature whose value to set

index - the index of the one-hot vector to set, as a string, which we will translate into a mapping

value - the value we want to set this one-hot index to

debugVector

public void debugVector(ConcatVector vector,
                        java.io.BufferedWriter bw)
                 throws java.io.IOException

This prints out a ConcatVector by mapping to the namespace, to make debugging learning algorithms easier.

Parameters:

vector - the vector to print

bw - the output stream to write to

Throws:

java.io.IOException