meka.core

Class ThresholdUtils

java.lang.Object

meka.core.ThresholdUtils

public abstract class ThresholdUtils
extends Object

ThresholdUtils - Helpful functions for calibrating thresholds.

Version:

March 2013

Author:

Jesse Read (jesse@tsc.uc3m.es)

Constructor Summary

Constructors

Constructor and Description
ThresholdUtils()

Method Summary

Modifier and Type	Method and Description
static double	calibrateThreshold(ArrayList<double[]> Y, double LC_train) CalibrateThreshold - Calibrate a threshold using PCut: the threshold which results in the best approximation of the label cardinality of the training set.
static double[]	calibrateThresholds(ArrayList<double[]> Y, double[] LC_train) CalibrateThreshold - Calibrate a vector of thresholds (one for each label) using PCut: the threshold t[j] which results in the best approximation of the frequency of the j-th label in the training data.
static int[][]	round(double[][] Rpred) Round - simply round numbers (e.g., 2.0 to 2) -- for multi-target data (where we don't *yet* use a threshold).
static int[][]	threshold(double[][] Rpred, double t) Threshold - returns the labels after the prediction-confidence vector is passed through threshold.
static int[][]	threshold(double[][] Rpred, double[] t) Threshold - returns the labels after the prediction-confidence vector is passed through a vector of thresholds.
static int[]	threshold(double[] rpred, String ts) Threshold - returns the labels after the prediction-confidence vector is passed through threshold(s).
static double[]	thresholdStringToArray(String top, int L) ThresholdStringToArray - parse a threshold option string to an array of L thresholds (one for each label variable).

Methods inherited from class java.lang.Object

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

Constructor Detail

ThresholdUtils

public ThresholdUtils()

Method Detail

thresholdStringToArray

public static double[] thresholdStringToArray(String top,
                                              int L)

ThresholdStringToArray - parse a threshold option string to an array of L thresholds (one for each label variable).

calibrateThreshold

public static double calibrateThreshold(ArrayList<double[]> Y,
                                        double LC_train)

CalibrateThreshold - Calibrate a threshold using PCut: the threshold which results in the best approximation of the label cardinality of the training set.

Parameters:

Y - labels

LC_train - label cardinality of the training set

calibrateThresholds

public static double[] calibrateThresholds(ArrayList<double[]> Y,
                                           double[] LC_train)

CalibrateThreshold - Calibrate a vector of thresholds (one for each label) using PCut: the threshold t[j] which results in the best approximation of the frequency of the j-th label in the training data.

Parameters:

Y - labels

LC_train - average frequency of each label

threshold

public static final int[][] threshold(double[][] Rpred,
                                      double[] t)

Threshold - returns the labels after the prediction-confidence vector is passed through a vector of thresholds.

Parameters:

Rpred - label confidence predictions in [0,1]

t - threshold for each label

threshold

public static final int[][] threshold(double[][] Rpred,
                                      double t)

Threshold - returns the labels after the prediction-confidence vector is passed through threshold.

Parameters:

Rpred - label confidence predictions in [0,1]

t - threshold

threshold

public static final int[] threshold(double[] rpred,
                                    String ts)

Threshold - returns the labels after the prediction-confidence vector is passed through threshold(s).

Parameters:

rpred - label confidence predictions in [0,1]

ts - threshold String

round

public static final int[][] round(double[][] Rpred)

Round - simply round numbers (e.g., 2.0 to 2) -- for multi-target data (where we don't *yet* use a threshold).

Parameters:

Rpred - class predictions in [0,1,...,K]

Returns:

integer representation of the predictions