org.openimaj.ml.linear.learner

Class IncrementalBilinearSparseOnlineLearner

java.lang.Object

org.openimaj.ml.linear.learner.IncrementalBilinearSparseOnlineLearner

All Implemented Interfaces:

OnlineLearner<Map<String,Map<String,Double>>,Map<String,Double>>

public class IncrementalBilinearSparseOnlineLearner
extends Object
implements OnlineLearner<Map<String,Map<String,Double>>,Map<String,Double>>

Author:

Sina Samangooei (ss@ecs.soton.ac.uk)

Constructor Summary

Constructors

Constructor and Description
IncrementalBilinearSparseOnlineLearner() Instantiates with the default params
IncrementalBilinearSparseOnlineLearner(BilinearLearnerParameters params)

Method Summary

Modifier and Type	Method and Description
Pair<gov.sandia.cognition.math.matrix.Matrix>	asMatrixPair(IndependentPair<Map<String,Map<String,Double>>,Map<String,Double>> in)
Pair<gov.sandia.cognition.math.matrix.Matrix>	asMatrixPair(IndependentPair<Map<String,Map<String,Double>>,Map<String,Double>> xy, int nfeatures, int nusers, int ntasks) Given a sparse pair of user/words and value construct a pair of matricies using the current mappings of words and users to matrix rows.
Pair<gov.sandia.cognition.math.matrix.Matrix>	asMatrixPair(Map<String,Map<String,Double>> x, Map<String,Double> y)
BilinearSparseOnlineLearner	getBilinearLearner()
BilinearSparseOnlineLearner	getBilinearLearner(int nusers, int nwords) Construct a learner with the desired number of users and words.
com.google.common.collect.BiMap<String,Integer>	getDependantValues()
BilinearLearnerParameters	getParams()
com.google.common.collect.BiMap<String,Integer>	getUsers()
com.google.common.collect.BiMap<String,Integer>	getVocabulary()
Map<String,Double>	predict(Map<String,Map<String,Double>> x)
void	process(Map<String,Map<String,Double>> x, Map<String,Double> y)
void	reinitParams()
void	updateUserValues(Map<String,Map<String,Double>> x, Map<String,Double> y) Update the incremental learner and underlying weight matricies to reflect potentially novel users, words and values to learn against

Methods inherited from class java.lang.Object

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

Constructor Detail

IncrementalBilinearSparseOnlineLearner

public IncrementalBilinearSparseOnlineLearner()

Instantiates with the default params

IncrementalBilinearSparseOnlineLearner

public IncrementalBilinearSparseOnlineLearner(BilinearLearnerParameters params)

Parameters:

params -

Method Detail

reinitParams

public void reinitParams()

getParams

public BilinearLearnerParameters getParams()

Returns:

the current parameters

process

public void process(Map<String,Map<String,Double>> x,
                    Map<String,Double> y)

Specified by:

process in interface OnlineLearner<Map<String,Map<String,Double>>,Map<String,Double>>

updateUserValues

public void updateUserValues(Map<String,Map<String,Double>> x,
                             Map<String,Double> y)

Update the incremental learner and underlying weight matricies to reflect potentially novel users, words and values to learn against

Parameters:

x -

y -

getBilinearLearner

public BilinearSparseOnlineLearner getBilinearLearner(int nusers,
                                                      int nwords)

Construct a learner with the desired number of users and words. If users and words beyond those in the current model are asked for their parameters are set to 0

Parameters:

nusers -

nwords -

Returns:

a new BilinearSparseOnlineLearner

getBilinearLearner

public BilinearSparseOnlineLearner getBilinearLearner()

Returns:

the underlying BilinearSparseOnlineLearner with the current number of users and words

asMatrixPair

public Pair<gov.sandia.cognition.math.matrix.Matrix> asMatrixPair(IndependentPair<Map<String,Map<String,Double>>,Map<String,Double>> xy,
                                                                  int nfeatures,
                                                                  int nusers,
                                                                  int ntasks)

Given a sparse pair of user/words and value construct a pair of matricies using the current mappings of words and users to matrix rows. Note: if users or words which have not yet be

Parameters:

xy -

nfeatures - the number of words total in the returned X matrix

nusers - the number of users total in the returned X matrix

ntasks - the number of tasks in the returned Y matrix

Returns:

the matrix pair representing the X and Y input

predict

public Map<String,Double> predict(Map<String,Map<String,Double>> x)

Specified by:

predict in interface OnlineLearner<Map<String,Map<String,Double>>,Map<String,Double>>

getVocabulary

public com.google.common.collect.BiMap<String,Integer> getVocabulary()

Returns:

the vocabulary

asMatrixPair

public Pair<gov.sandia.cognition.math.matrix.Matrix> asMatrixPair(IndependentPair<Map<String,Map<String,Double>>,Map<String,Double>> in)

Parameters:

in -

Returns:

calls asMatrixPair(IndependentPair, int, int, int) with the current number of words, users and value

asMatrixPair

public Pair<gov.sandia.cognition.math.matrix.Matrix> asMatrixPair(Map<String,Map<String,Double>> x,
                                                                  Map<String,Double> y)

Parameters:

x -

y -

Returns:

calls asMatrixPair(IndependentPair, int, int, int) with the current number of words, users and value

getDependantValues

public com.google.common.collect.BiMap<String,Integer> getDependantValues()

Returns:

the current map of dependent values to indexes

getUsers

public com.google.common.collect.BiMap<String,Integer> getUsers()