Class ImprovedRiveretTheoryLearner

java.lang.Object
org.tweetyproject.arg.dung.learning.ImprovedRiveretTheoryLearner

public class ImprovedRiveretTheoryLearner extends Object
Implementation of the algorithm for learning (grounded) labelings from with some improvements: Riveret, RĂ©gis, and Guido Governatori. "On learning attacks in probabilistic abstract argumentation." 2016. - improves result for argumentation frameworks with self-attacking arguments - instead of stopping when no undecided attacks are left, we always use the full computational budget - added additional parameter for pruning instead of pruning at 0. eg budget: 500 threshold: -10 - added fifth rule to capture self-attacking arguments - extended rule 3 - adjusted weight updates per rule
Author:
Lars Bengel
  • Constructor Details

    • ImprovedRiveretTheoryLearner

      public ImprovedRiveretTheoryLearner(Collection<Argument> arguments, int max_cycles)
      initialize learner for the given set of arguments
      Parameters:
      arguments - a set of arguments
      max_cycles - the maximal number of cycles
  • Method Details

    • showWeights

      public void showWeights()
      show Weights
    • learnLabelings

      public DungTheory learnLabelings(ArrayList<Labeling> labelings, boolean prune, int threshold)
      learn random labelings from the given List until no undecided attacks are left in the theory
      Parameters:
      labelings - a list of labelings
      prune - if true, remove discarded attacks after each step
      threshold - the treshold
      Returns:
      the learned dung theory
    • learnLabelings

      public DungTheory learnLabelings(ArrayList<Labeling> labelings)
      learn theory without pruning discarded attacks
      Parameters:
      labelings - a list of labelings
      Returns:
      the learned dung theory