Class LearnCBJ

  • All Implemented Interfaces:
    Learn
    Direct Known Subclasses:
    LearnDBT

    public class LearnCBJ
    extends LearnExplained
    Conflict-based Backjumping[1] (CBJ) explanation strategy. It backtracks up to most recent decision involved in the explanation, and forget younger decisions.

    [1]: P. Prosser, Hybrid algorithms for the constraint satisfaction problem, Computational Intelligence (93).

    Created by cprudhom on 02/09/15. Project: choco.

    Author:
    Charles Prud'homme, Narendra Jussien
    • Constructor Detail

      • LearnCBJ

        public LearnCBJ​(Model mModel,
                        boolean nogoodFromConflict,
                        boolean userFeedbackOn)
        Create a Conflict-based Backjumping strategy.
        Parameters:
        mModel - the solver to instrument
        nogoodFromConflict - set to true to extract nogoods from explanations.
        userFeedbackOn - set to true to record causes in explanations (required for user feedback mainly).
    • Method Detail

      • identifyRefutedDecision

        protected void identifyRefutedDecision​(int nworld)
        Identify the decision to reconsider, and explain its refutation in the explanation data base
        Parameters:
        nworld - index of the world to backtrack to