Classifier (Interface Manipulation Package)

[using it]

Interface Manipulation Package v4.0 (2008-06-25)

©copyright

imp.classifier
Class Classifier

java.lang.Object
  imp.classifier.Classifier

Direct Known Subclasses:: SvmClassifier, YgClassifier, YgPolynomialClassifier

public class Classifier
extends Object
extends Object

Defines a numerical data classifier.

Formally this may be defined as a function: double[dim] -> String which associated to a numerical value vector a category given as a string.

Very easily a user construct a classifier and:

add calibration data of known category to define this transformation,
getCategory the category of new data samples when the training is performed,

as defined by the proposed methods.

As a basic mechanism a nearest-neighbor classifier is implemented

Classifiers based on this interface should overwrite:

train methods to react to a new calibration sample input,
getIndex method to returns the category of a given data,

Nested Class Summary
`static class`	`Classifier.Data` Defines a classifier data set.

Constructor Summary
`Classifier()` Constructs a raw-classifier (nearest-neighbor classifier).

Method Summary
`Classifier`	`add(double[] data, String category, double prob)` Add a data calibration sample and return this.
`String`	`getCategory(double[] data)` Get the category for a given input.
`Classifier.Data`	`getData()` Returns the current calibration data set.
`int`	`getIndex(double[] data)` Performs a classification This method is called when a data category is requested.
`double`	`getMargin(double[] data, String category)` Estimates the margin of a given sample.
`String`	`getName(int index)` Gets the name of a category.
`String`	`toString()` Returns a string view of this classifier as a XML text with some classifier properties.
`void`	`train(Classifier.Data data)` Performs the classifier training This method is automatically called, after calibration samples have been added and before the next data category request.

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

Constructor Detail

Classifier

public Classifier()

Constructs a raw-classifier (nearest-neighbor classifier).

Method Detail

add

public final Classifier add(double[] data,
                            String category,
                            double prob)

Add a data calibration sample and return this.

Parameters:: data - calibration input value.; category - [optional, default is ""] category name or "" if the category is undefined.; prob - [optional, default is 1] Sample precision probability: 1 for an exact sample, 0 for an undefined sample.

getCategory

public final String getCategory(double[] data)

Get the category for a given input.

Parameters:: data - data input value.
Returns:: The category name or "" if no classification available.

getData

public Classifier.Data getData()

Returns the current calibration data set.

getIndex

public int getIndex(double[] data)

Performs a classification

This method is called when a data category is requested.

Parameters:: data - data input value.
Returns:: The category index or -1 if no classification available.

getMargin

public double getMargin(double[] data,
                        String category)

Estimates the margin of a given sample.

Parameters:: data - sample data.; category - sample category.
Returns:: The sample margin estimate (i.e. the distance from this sample to a the closest sample of another category after which the classification fail)

getName

public String getName(int index)

Gets the name of a category.

Parameters:: index - The category index.
Returns:: The category name.

toString

public String toString()

Returns a string view of this classifier as a XML text with some classifier properties.

Overrides:: toString in class Object

train

public void train(Classifier.Data data)

Performs the classifier training

This method is automatically called, after calibration samples have been added and before the next data category request.

Parameters:: data - [optional, default is the current calibration data set] The calibration data set to be used to train the classifier.

imp.classifier Class Classifier