jsat.classifiers.calibration

Class IsotonicCalibration

java.lang.Object

jsat.classifiers.calibration.BinaryCalibration

jsat.classifiers.calibration.IsotonicCalibration

All Implemented Interfaces:

Serializable, Cloneable, Classifier, Parameterized

public class IsotonicCalibration
extends BinaryCalibration

Isotonic Calibration is non-parametric, and only assumes that the underlying distribution from negative to positive examples is strictly a non-decreasing function. It will then attempt to model the distribution. This may over-fit for small data sizes, and imposes an additional O(log n) search look up when performing classification, where n is <= the number of data points in the data set.

Isotonic Calibration inherently creates non-adjacent bins of varying size. Smooth transitions in output probability are created by simple linear interpolation between bin values.

See: Niculescu-Mizil, A.,&Caruana, R. (2005). Predicting Good Probabilities with Supervised Learning. International Conference on Machine Learning (pp. 625–632). Retrieved from here

Author:

Edward Raff

See Also:

Serialized Form

Nested Class Summary

Nested classes/interfaces inherited from class jsat.classifiers.calibration.BinaryCalibration

BinaryCalibration.CalibrationMode

Field Summary

Fields inherited from class jsat.classifiers.calibration.BinaryCalibration

base, folds, holdOut, mode

Constructor Summary

Constructors

Modifier	Constructor and Description
	IsotonicCalibration(BinaryScoreClassifier base, BinaryCalibration.CalibrationMode mode) Creates a new Isotonic Calibration object
protected	IsotonicCalibration(IsotonicCalibration toCopy) Copy constructor

Method Summary

Modifier and Type	Method and Description
protected void	calibrate(boolean[] label, double[] deci, int len) This method perform the model calibration on the outputs verse the class labels.
CategoricalResults	classify(DataPoint data) Performs classification on the given data point.
IsotonicCalibration	clone()
boolean	supportsWeightedData() Indicates whether the model knows how to train using weighted data points.

Methods inherited from class jsat.classifiers.calibration.BinaryCalibration

getCalibrationFolds, getCalibrationHoldOut, getCalibrationMode, getParameter, getParameters, setCalibrationFolds, setCalibrationHoldOut, setCalibrationMode, trainC, trainC

Methods inherited from class java.lang.Object

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

Constructor Detail

IsotonicCalibration

public IsotonicCalibration(BinaryScoreClassifier base,
                           BinaryCalibration.CalibrationMode mode)

Creates a new Isotonic Calibration object

Parameters:

base - the base model to calibrate the outputs of

mode - the calibration mode to use

IsotonicCalibration

protected IsotonicCalibration(IsotonicCalibration toCopy)

Copy constructor

Parameters:

toCopy - the object to copy

Method Detail

calibrate

protected void calibrate(boolean[] label,
                         double[] deci,
                         int len)

Description copied from class: BinaryCalibration

This method perform the model calibration on the outputs verse the class labels.

Specified by:

calibrate in class BinaryCalibration

Parameters:

label - the set of labels, where true indicates the positive class label, and false indicates the negative class label.

deci - the score associated with each label from the learning algorithm.

len - the number of values (from zero) of the label and scores array to use. This value may be less than the actual array size

clone

public IsotonicCalibration clone()

Specified by:

clone in interface Classifier

Specified by:

clone in class BinaryCalibration

classify

public CategoricalResults classify(DataPoint data)

Description copied from interface: Classifier

Performs classification on the given data point.

Parameters:

data - the data point to classify

Returns:

the results of the classification.

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.

Returns:

true if the model supports weighted data, false otherwise