LVQ (Java Statistical Analysis Tool 0.0.8 API)

java.lang.Object
- jsat.classifiers.neuralnetwork.LVQ

All Implemented Interfaces:

Serializable, Cloneable, Classifier, Parameterized

Direct Known Subclasses:

LVQLLC
```
public class LVQ
extends Object
implements Classifier, Parameterized
```
Learning Vector Quantization (LVQ) is an algorithm that extends SOM to take advantage of label information to perform classification. It creates a number of representatives, or learning vectors, for each class. The LVs are then updated iteratively to push away from the wrong class and pull closer to the correct class. LVQ is equivalent to a type of 2 layer neural network.

Author:

Edward Raff

See Also:

Serialized Form

Nested Class Summary

Nested Classes
Modifier and Type	Class and Description
`static class`	`LVQ.LVQVersion` There are several LVQ versions, each one adding an additional case in which two LVs instead of one can be updated.

Field Summary

Fields
Modifier and Type	Field and Description
`static double`	`DEFAULT_EPS` The default eps distance factor between the two wining vectors 0.3
`static int`	`DEFAULT_ITERATIONS` The default number of iterations is 200
`static double`	`DEFAULT_LEARNING_RATE` The default learning rate 0.1
`static LVQ.LVQVersion`	`DEFAULT_LVQ_METHOD` The default method of LVQ to use LVQ3
`static double`	`DEFAULT_MSCALE` The default scaling factor for the `LVQ.LVQVersion.LVQ3` case is 0.30000000000000004
`static int`	`DEFAULT_REPS_PER_CLASS` The default number of representatives per class is 3
`static SeedSelectionMethods.SeedSelection`	`DEFAULT_SEED_SELECTION` The default seed selection method is SeedSelection.KPP
`static double`	`DEFAULT_STOPPING_DIST` The default stopping distance for convergence is 0.001
`protected DistanceMetric`	`dm` The distance metric to use
`protected VectorCollection<VecPaired<Vec,Integer>>`	`vc` Contains the Learning vectors paired with their index in the weights array
`protected int[]`	`weightClass` Array of the class that each learning vector represents
`protected Vec[]`	`weights` Array containing the learning vectors
`protected int[]`	`wins` Records the number of times each neuron won and was off the correct class during training.

Constructor Summary

Constructors
Modifier	Constructor and Description
	`LVQ(DistanceMetric dm, int iterations)` Creates a new LVQ instance
	`LVQ(DistanceMetric dm, int iterations, double learningRate, int representativesPerClass)` Creates a new LVQ instance
	`LVQ(DistanceMetric dm, int iterations, double learningRate, int representativesPerClass, LVQ.LVQVersion lvqVersion, DecayRate learningDecay)` Creates a new LVQ instance
`protected`	`LVQ(LVQ toCopy)` Copy Constructor

Method Summary

All Methods Instance Methods Concrete Methods
Modifier and Type	Method and Description
`CategoricalResults`	`classify(DataPoint data)` Performs classification on the given data point.
`LVQ`	`clone()`
`protected boolean`	`epsClose(double minDist, double minDist2)` Returns true if the two distance values are within an acceptable epsilon ratio of each other.
`DistanceMetric`	`getDistanceMetric()` Returns the distance metric to use
`double`	`getEpsilonDistance()` Sets the epsilon scale distance between learning vectors that may be allowed to two at a time.
`int`	`getIterations()` Returns the number of iterations of the algorithm to apply
`DecayRate`	`getLearningDecay()` Returns the method used to decay the learning rate over each iteration
`double`	`getLearningRate()` Returns the learning rate at which to apply updates during the algorithm.
`LVQ.LVQVersion`	`getLVQMethod()` Returns the version of the LVQ algorithm to use.
`double`	`getMScale()` Returns the scale used for the LVQ 3 learning algorithm update set.
`Parameter`	`getParameter(String paramName)` Returns the parameter with the given name.
`List<Parameter>`	`getParameters()` Returns the list of parameters that can be altered for this learner.
`int`	`getRepresentativesPerClass()` Returns the number of representatives to create for each class.
`SeedSelectionMethods.SeedSelection`	`getSeedSelection()` Returns the method of seed selection used
`double`	`getStoppingDist()` Returns the stopping distance used to terminate the algorithm early
`void`	`setDistanceMetric(DistanceMetric dm)` Sets the distance used for learning
`void`	`setEpsilonDistance(double eps)` Sets the epsilon multiplier that controls the maximum distance two learning vectors can be from each other in order to be updated at the same time.
`void`	`setIterations(int iterations)` Sets the number of learning iterations that will occur.
`void`	`setLearningDecay(DecayRate learningDecay)` Sets the decay rate to apply to the learning rate.
`void`	`setLearningRate(double learningRate)` Sets the learning rate of the algorithm.
`void`	`setLVQMethod(LVQ.LVQVersion lvqMethod)` Sets the version of LVQ used.
`void`	`setMScale(double mScale)` When using `LVQ.LVQVersion.LVQ3`, a 3rd case exists where up to two learning vectors can be updated at the same time if they have the same class.
`void`	`setRepresentativesPerClass(int representativesPerClass)` Sets the number of representatives to create for each class.
`void`	`setSeedSelection(SeedSelectionMethods.SeedSelection seedSelection)` Sets the seed selection method used to select the initial learning vectors
`void`	`setStoppingDist(double stoppingDist)` The algorithm terminates early if the learning vectors are only moving small distances.
`void`	`setVecCollectionFactory(VectorCollectionFactory<VecPaired<Vec,Integer>> vcf)` Sets the vector collection factory to use when storing the final learning vectors
`boolean`	`supportsWeightedData()` Indicates whether the model knows how to train using weighted data points.
`void`	`trainC(ClassificationDataSet dataSet)` Trains the classifier and constructs a model for classification using the given data set.
`void`	`trainC(ClassificationDataSet dataSet, ExecutorService threadPool)` Trains the classifier and constructs a model for classification using the given data set.

Methods inherited from class java.lang.Object
equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

- Field Detail
  - DEFAULT_ITERATIONS
```
public static final int DEFAULT_ITERATIONS
```
    The default number of iterations is 200
    
    See Also:
    
    Constant Field Values
  - DEFAULT_LEARNING_RATE
```
public static final double DEFAULT_LEARNING_RATE
```
    The default learning rate 0.1
    
    See Also:
    
    Constant Field Values
  - DEFAULT_EPS
```
public static final double DEFAULT_EPS
```
    The default eps distance factor between the two wining vectors 0.3
    
    See Also:
    
    Constant Field Values
  - DEFAULT_MSCALE
```
public static final double DEFAULT_MSCALE
```
    The default scaling factor for the LVQ.LVQVersion.LVQ3 case is 0.30000000000000004
    
    See Also:
    
    Constant Field Values
  - DEFAULT_LVQ_METHOD
```
public static final LVQ.LVQVersion DEFAULT_LVQ_METHOD
```
    The default method of LVQ to use LVQ3
  - DEFAULT_REPS_PER_CLASS
```
public static final int DEFAULT_REPS_PER_CLASS
```
    The default number of representatives per class is 3
    
    See Also:
    
    Constant Field Values
  - DEFAULT_STOPPING_DIST
```
public static final double DEFAULT_STOPPING_DIST
```
    The default stopping distance for convergence is 0.001
    
    See Also:
    
    Constant Field Values
  - DEFAULT_SEED_SELECTION
```
public static final SeedSelectionMethods.SeedSelection DEFAULT_SEED_SELECTION
```
    The default seed selection method is SeedSelection.KPP
  - dm
```
protected DistanceMetric dm
```
    The distance metric to use
  - weights
```
protected Vec[] weights
```
    Array containing the learning vectors
  - weightClass
```
protected int[] weightClass
```
    Array of the class that each learning vector represents
  - wins
```
protected int[] wins
```
    Records the number of times each neuron won and was off the correct class during training. Neurons that end with a count of zero wins will be ignored
  - vc
```
protected VectorCollection<VecPaired<Vec,Integer>> vc
```
    Contains the Learning vectors paired with their index in the weights array
- Constructor Detail
  - LVQ
```
public LVQ(DistanceMetric dm,
           int iterations)
```
    Creates a new LVQ instance
    
    Parameters:
    
    dm - the distance metric to use
    
    iterations - the number of iterations to perform
  - LVQ
```
public LVQ(DistanceMetric dm,
           int iterations,
           double learningRate,
           int representativesPerClass)
```
    Creates a new LVQ instance
    
    Parameters:
    
    dm - the distance metric to use
    
    iterations - the number of iterations to perform
    
    learningRate - the learning rate to use when updating
    
    representativesPerClass - the number of representatives to create for each class
  - LVQ
```
public LVQ(DistanceMetric dm,
           int iterations,
           double learningRate,
           int representativesPerClass,
           LVQ.LVQVersion lvqVersion,
           DecayRate learningDecay)
```
    Creates a new LVQ instance
    
    Parameters:
    
    dm - the distance metric to use
    
    iterations - the number of iterations to perform
    
    learningRate - the learning rate to use when updating
    
    representativesPerClass - the number of representatives to create for each class
    
    lvqVersion - the version of LVQ to use
    
    learningDecay - the amount of decay to apply to the learning rate
  - LVQ
```
protected LVQ(LVQ toCopy)
```
    Copy Constructor
    
    Parameters:
    
    toCopy - version to copy
- Method Detail
  - setMScale
```
public void setMScale(double mScale)
```
    When using LVQ.LVQVersion.LVQ3, a 3rd case exists where up to two learning vectors can be updated at the same time if they have the same class. To avoid over fitting, an additional regularizing weight is placed upon the learning rate for their update. THis sets the additional multiplied. It is suggested to use a value in the range of [0.1, 0.5]
    
    Parameters:
    
    mScale - the multiplication factor to apply to the learning vectors
  - getMScale
```
public double getMScale()
```
    Returns the scale used for the LVQ 3 learning algorithm update set.
    
    Returns:
    
    a scale used during LVQ3
  - setEpsilonDistance
```
public void setEpsilonDistance(double eps)
```
    Sets the epsilon multiplier that controls the maximum distance two learning vectors can be from each other in order to be updated at the same time. If they are too far apart, only one can be updated. It is recommended to use a value in the range [0.1, 0.3]
    
    Parameters:
    
    eps - the scale factor of the maximum distance for two learning vectors to be updated at the same time
  - getEpsilonDistance
```
public double getEpsilonDistance()
```
    Sets the epsilon scale distance between learning vectors that may be allowed to two at a time.
    
    Returns:
    
    the scale of the allowable distance between learning vectors when updating
  - setLearningRate
```
public void setLearningRate(double learningRate)
```
    Sets the learning rate of the algorithm. It should be set in accordance with setLearningDecay(jsat.math.decayrates.DecayRate).
    
    Parameters:
    
    learningRate - the learning rate to use
  - getLearningRate
```
public double getLearningRate()
```
    Returns the learning rate at which to apply updates during the algorithm.
    
    Returns:
    
    the learning rate to use
  - setLearningDecay
```
public void setLearningDecay(DecayRate learningDecay)
```
    Sets the decay rate to apply to the learning rate.
    
    Parameters:
    
    learningDecay - the rate to decay the learning rate
  - getLearningDecay
```
public DecayRate getLearningDecay()
```
    Returns the method used to decay the learning rate over each iteration
    
    Returns:
    
    the decay rate used at each iteration
  - setIterations
```
public void setIterations(int iterations)
```
    Sets the number of learning iterations that will occur.
    
    Parameters:
    
    iterations - the number of iterations for the algorithm to use
  - getIterations
```
public int getIterations()
```
    Returns the number of iterations of the algorithm to apply
    
    Returns:
    
    the number of iterations to perform
  - setRepresentativesPerClass
```
public void setRepresentativesPerClass(int representativesPerClass)
```
    Sets the number of representatives to create for each class. It is possible to have an unbalanced number of representatives per class, but that is not currently supported. Increasing the number of representatives per class increases the complexity of the decision boundary that can be learned.
    
    Parameters:
    
    representativesPerClass - the number of representatives to create for each class
  - getRepresentativesPerClass
```
public int getRepresentativesPerClass()
```
    Returns the number of representatives to create for each class.
    
    Returns:
    
    the number of representatives to create for each class.
  - setLVQMethod
```
public void setLVQMethod(LVQ.LVQVersion lvqMethod)
```
    Sets the version of LVQ used.
    
    Parameters:
    
    lvqMethod - the version of LVQ to use
  - getLVQMethod
```
public LVQ.LVQVersion getLVQMethod()
```
    Returns the version of the LVQ algorithm to use.
    
    Returns:
    
    the version of the LVQ algorithm to use.
  - setDistanceMetric
```
public void setDistanceMetric(DistanceMetric dm)
```
    Sets the distance used for learning
    
    Parameters:
    
    dm - the distance metric to use
  - getDistanceMetric
```
public DistanceMetric getDistanceMetric()
```
    Returns the distance metric to use
    
    Returns:
    
    the distance metric to use
  - setStoppingDist
```
public void setStoppingDist(double stoppingDist)
```
    The algorithm terminates early if the learning vectors are only moving small distances. The stopping distance is the minimum distance that one of the learning vectors must move for the algorithm to continue.
    
    Parameters:
    
    stoppingDist - the minimum distance for each learning vector to move
  - getStoppingDist
```
public double getStoppingDist()
```
    Returns the stopping distance used to terminate the algorithm early
    
    Returns:
    
    the stopping distance used toe nd the algorithm early
  - setSeedSelection
```
public void setSeedSelection(SeedSelectionMethods.SeedSelection seedSelection)
```
    Sets the seed selection method used to select the initial learning vectors
    
    Parameters:
    
    seedSelection - the method of initialing LVQ
  - getSeedSelection
```
public SeedSelectionMethods.SeedSelection getSeedSelection()
```
    Returns the method of seed selection used
    
    Returns:
    
    the method of seed selection used
  - getParameters
```
public List<Parameter> getParameters()
```
    Description copied from interface: Parameterized
    
    Returns the list of parameters that can be altered for this learner.
    
    Specified by:
    
    getParameters in interface Parameterized
    
    Returns:
    
    the list of parameters that can be altered for this learner.
  - getParameter
```
public Parameter getParameter(String paramName)
```
    Description copied from interface: Parameterized
    
    Returns the parameter with the given name. Two different strings may map to a single Parameter object. An ASCII only string, and a Unicode style string.
    
    Specified by:
    
    getParameter in interface Parameterized
    
    Parameters:
    
    paramName - the name of the parameter to obtain
    
    Returns:
    
    the Parameter in question, or null if no such named Parameter exists.
  - setVecCollectionFactory
```
public void setVecCollectionFactory(VectorCollectionFactory<VecPaired<Vec,Integer>> vcf)
```
    Sets the vector collection factory to use when storing the final learning vectors
    
    Parameters:
    
    vcf - the vector collection factory to use
  - classify
```
public CategoricalResults classify(DataPoint data)
```
    Description copied from interface: Classifier
    
    Performs classification on the given data point.
    
    Specified by:
    
    classify in interface Classifier
    
    Parameters:
    
    data - the data point to classify
    
    Returns:
    
    the results of the classification.
  - epsClose
```
protected boolean epsClose(double minDist,
                           double minDist2)
```
    Returns true if the two distance values are within an acceptable epsilon ratio of each other.
    
    Parameters:
    
    minDist - the first distance
    
    minDist2 - the second distance
    
    Returns:
    
    true if the are acceptable close
  - trainC
```
public void trainC(ClassificationDataSet dataSet,
                   ExecutorService threadPool)
```
    Description copied from interface: Classifier
    
    Trains the classifier and constructs a model for classification using the given data set. If the training method knows how, it will used the threadPool to conduct training in parallel. This method will block until the training has completed.
    
    Specified by:
    
    trainC in interface Classifier
    
    Parameters:
    
    dataSet - the data set to train on
    
    threadPool - the source of threads to use.
  - trainC
```
public void trainC(ClassificationDataSet dataSet)
```
    Description copied from interface: Classifier
    
    Trains the classifier and constructs a model for classification using the given data set.
    
    Specified by:
    
    trainC in interface Classifier
    
    Parameters:
    
    dataSet - the data set to train on
  - supportsWeightedData
```
public boolean supportsWeightedData()
```
    Description copied from interface: Classifier
    
    Indicates whether the model knows how to train using weighted data points. If it does, the model will train assuming the weights. The values returned by this method may change depending on the parameters set for the model.
    
    Specified by:
    
    supportsWeightedData in interface Classifier
    
    Returns:
    
    true if the model supports weighted data, false otherwise
  - clone
```
public LVQ clone()
```
    Specified by:
    
    clone in interface Classifier
    
    Overrides:
    
    clone in class Object

Class LVQ

Nested Class Summary

Field Summary

Constructor Summary

Method Summary

Methods inherited from class java.lang.Object

Field Detail

DEFAULT_ITERATIONS

DEFAULT_LEARNING_RATE

DEFAULT_EPS

DEFAULT_MSCALE

DEFAULT_LVQ_METHOD

DEFAULT_REPS_PER_CLASS

DEFAULT_STOPPING_DIST

DEFAULT_SEED_SELECTION

dm

weights

weightClass

wins

vc

Constructor Detail

LVQ

LVQ

LVQ

LVQ

Method Detail

setMScale

getMScale

setEpsilonDistance

getEpsilonDistance

setLearningRate

getLearningRate

setLearningDecay

getLearningDecay

setIterations

getIterations

setRepresentativesPerClass

getRepresentativesPerClass

setLVQMethod

getLVQMethod

setDistanceMetric

getDistanceMetric

setStoppingDist

getStoppingDist

setSeedSelection

getSeedSelection

getParameters

getParameter

setVecCollectionFactory

classify

epsClose

trainC

trainC

supportsWeightedData

clone