calhoun.analysis.crf.features.tricycle13
Class KmerFeatures

java.lang.Object
  calhoun.analysis.crf.AbstractFeatureManager<java.lang.Character>
      calhoun.analysis.crf.features.tricycle13.KmerFeatures

All Implemented Interfaces:

FeatureManager<java.lang.Character>, FeatureManagerNode<java.lang.Character>, java.io.Serializable

public class KmerFeatures
extends AbstractFeatureManager<java.lang.Character>
implements FeatureManagerNode<java.lang.Character>

trains on the data and then evaluates to P(state | label) for given Kmers. Used for historical reasons. Emission markov generally does better.

See Also:

Serialized Form

Nested Class Summary

static class

KmerFeatures.Cardinality

Constructor Summary

KmerFeatures()

KmerFeatures(KmerFeatures.Cardinality cardinality)

KmerFeatures(java.util.List<int[]> kmerDefs)

KmerFeatures(java.util.List<int[]> kmerDefs, KmerFeatures.Cardinality cardinality)

Method Summary

void	evaluateNode(InputSequence<? extends java.lang.Character> seq, int pos, int state, FeatureList result) Evaluates the set of features managed by this object for the given arguments.
java.lang.String	getFeatureName(int featureIndex) Returns a human identifiable name for the feature referenced by a given index.
java.lang.String	getKmer(InputSequence<? extends java.lang.Character> seq, int pos, int[] def)
double	getKmerProb(int kmerIndex, java.lang.String kmer, int label) Returns an individual entry from the counts list.
int	getNumFeatures() Returns the number of features maintained by this FeatureManager.
java.lang.String	kmerName(int index) Returns a string representation of a given kmer definition
void	setKmerDefinitions(java.util.List<java.util.List<java.lang.Integer>> defs)
void	setRareThreshold(int threshold)
void	train(int startingIndex, ModelManager modelInfo, java.util.List<? extends TrainingSequence<? extends java.lang.Character>> data) Computes the P(label | kmer) for each kmer across all of the training data.

Methods inherited from class calhoun.analysis.crf.AbstractFeatureManager

getCacheStrategy, getInputComponent, setInputComponent

Methods inherited from class java.lang.Object

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

Methods inherited from interface calhoun.analysis.crf.FeatureManager

getCacheStrategy, getInputComponent, setInputComponent

Constructor Detail

KmerFeatures

public KmerFeatures(KmerFeatures.Cardinality cardinality)

KmerFeatures

public KmerFeatures()

KmerFeatures

public KmerFeatures(java.util.List<int[]> kmerDefs)

KmerFeatures

public KmerFeatures(java.util.List<int[]> kmerDefs,
                    KmerFeatures.Cardinality cardinality)

Method Detail

setKmerDefinitions

public void setKmerDefinitions(java.util.List<java.util.List<java.lang.Integer>> defs)

setRareThreshold

public void setRareThreshold(int threshold)

getNumFeatures

public int getNumFeatures()

Description copied from interface: FeatureManager

Returns the number of features maintained by this FeatureManager. This number must be fixed after the call to trainFeatures is complete.

Specified by:

getNumFeatures in interface FeatureManager<java.lang.Character>

Returns:

number of features managed by this FeatureManager

getFeatureName

public java.lang.String getFeatureName(int featureIndex)

Description copied from interface: FeatureManager

Returns a human identifiable name for the feature referenced by a given index. Used for display purposes only.

Specified by:

getFeatureName in interface FeatureManager<java.lang.Character>

Parameters:

featureIndex - the index of this feature

Returns:

the human readable name of this feature

evaluateNode

public void evaluateNode(InputSequence<? extends java.lang.Character> seq,
                         int pos,
                         int state,
                         FeatureList result)

Description copied from interface: FeatureManagerNode

Evaluates the set of features managed by this object for the given arguments.

Specified by:

evaluateNode in interface FeatureManagerNode<java.lang.Character>

getKmer

public java.lang.String getKmer(InputSequence<? extends java.lang.Character> seq,
                                int pos,
                                int[] def)

kmerName

public java.lang.String kmerName(int index)

Returns a string representation of a given kmer definition

getKmerProb

public double getKmerProb(int kmerIndex,
                          java.lang.String kmer,
                          int label)

Returns an individual entry from the counts list.

train

public void train(int startingIndex,
                  ModelManager modelInfo,
                  java.util.List<? extends TrainingSequence<? extends java.lang.Character>> data)

Computes the P(label | kmer) for each kmer across all of the training data. These will used as features values.

Specified by:

train in interface FeatureManager<java.lang.Character>

Parameters:

startingIndex - the feature index of the first feature owned by this FeatureManager. Each FeatureManager must use up consecutive indexes, so the last index used will be startingIndex + numFeatures - 1.

modelInfo - the model that contains this feature

data - the full list of training sequences to use to train the feature