ai.libs.jaicore.ml.dyadranking.activelearning

Class PrototypicalPoolBasedActiveDyadRanker

java.lang.Object

ai.libs.jaicore.ml.dyadranking.activelearning.ActiveDyadRanker

ai.libs.jaicore.ml.dyadranking.activelearning.ARandomlyInitializingDyadRanker

ai.libs.jaicore.ml.dyadranking.activelearning.PrototypicalPoolBasedActiveDyadRanker

public class PrototypicalPoolBasedActiveDyadRanker
extends ARandomlyInitializingDyadRanker

A prototypical active dyad ranker based on the idea of uncertainty sampling. First a constant number of random queries is sampled at the beginning. Then the sampling strategy randomly selects a problem instance in each query step. Afterwards it selects those two alternatives for pairwise comparison, for which the difference of the skill values is minimal, as these are the pairs the Plackett Luce model is least certain about. This procedure is repeated a constant number of times to create a minibatch for updating the model.

Field Summary

Fields inherited from class ai.libs.jaicore.ml.dyadranking.activelearning.ActiveDyadRanker

poolProvider, ranker

Constructor Summary

Constructors

Constructor and Description
PrototypicalPoolBasedActiveDyadRanker(PLNetDyadRanker ranker, IDyadRankingPoolProvider poolProvider, int maxBatchSize, int lengthOfTopRankingToConsider, double ratioOfOldInstancesForMinibatch, int numberRandomQueriesAtStart, int seed)

Method Summary

Modifier and Type	Method and Description
void	activelyTrainWithOneInstance()
int	getLengthOfTopRankingToConsider()
double	getRatioOfOldInstancesForMinibatch()
void	setLengthOfTopRankingToConsider(int lengthOfTopRankingToConsider)
void	setRatioOfOldInstancesForMinibatch(double ratioOfOldInstancesForMinibatch)

Methods inherited from class ai.libs.jaicore.ml.dyadranking.activelearning.ARandomlyInitializingDyadRanker

activelyTrain, getDyadStats, getInstanceFeatures, getIteration, getMinibatchSize, getNumberRandomQueriesAtStart, getRandom, updateRanker

Methods inherited from class ai.libs.jaicore.ml.dyadranking.activelearning.ActiveDyadRanker

getPoolProvider, getRanker, setPoolProvider, setRanker

Methods inherited from class java.lang.Object

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

Constructor Detail

PrototypicalPoolBasedActiveDyadRanker

public PrototypicalPoolBasedActiveDyadRanker(PLNetDyadRanker ranker,
                                             IDyadRankingPoolProvider poolProvider,
                                             int maxBatchSize,
                                             int lengthOfTopRankingToConsider,
                                             double ratioOfOldInstancesForMinibatch,
                                             int numberRandomQueriesAtStart,
                                             int seed)

Method Detail

activelyTrainWithOneInstance

public void activelyTrainWithOneInstance()
                                  throws TrainingException

Specified by:

activelyTrainWithOneInstance in class ARandomlyInitializingDyadRanker

Throws:

TrainingException

getRatioOfOldInstancesForMinibatch

public double getRatioOfOldInstancesForMinibatch()

setRatioOfOldInstancesForMinibatch

public void setRatioOfOldInstancesForMinibatch(double ratioOfOldInstancesForMinibatch)

getLengthOfTopRankingToConsider

public int getLengthOfTopRankingToConsider()

setLengthOfTopRankingToConsider

public void setLengthOfTopRankingToConsider(int lengthOfTopRankingToConsider)