ai.libs.jaicore.ml.dyadranking.loss

Class KendallsTauDyadRankingLoss

java.lang.Object

ai.libs.jaicore.ml.dyadranking.loss.KendallsTauDyadRankingLoss

All Implemented Interfaces:

DyadRankingLossFunction

public class KendallsTauDyadRankingLoss
extends java.lang.Object
implements DyadRankingLossFunction

Computes the rank correlation measure known as Kendall's tau coefficient, i.e. (C - D) / (K * (K-1) /2), where C and D are the number of concordant (put in the right order) and discordant (put in the wrong order) pairs of dyads and K is the length of the dyad ranking. Lies between -1 (reversed order) and +1 (same order). Assumes the dyads in the ranking to be pairwise distinct.

Constructor Summary

Constructors

Constructor and Description
KendallsTauDyadRankingLoss()

Method Summary

Modifier and Type	Method and Description
double	loss(IDyadRankingInstance actual, IDyadRankingInstance predicted) Computes the loss between the actual dyad ordering and predicted dyad ordering, represented by dyad ranking instances.

Methods inherited from class java.lang.Object

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

Constructor Detail

KendallsTauDyadRankingLoss

public KendallsTauDyadRankingLoss()

Method Detail

loss

public double loss(IDyadRankingInstance actual,
                   IDyadRankingInstance predicted)

Description copied from interface: DyadRankingLossFunction

Computes the loss between the actual dyad ordering and predicted dyad ordering, represented by dyad ranking instances.

Specified by:

loss in interface DyadRankingLossFunction

Parameters:

actual - the correct ordering

predicted - the predicted ordering

Returns:

the loss between the predicted and correct ordering, depending on the implementation