Answer ( 1 )

    This answer is edited.

    Benefits of Dempster-Shafer Theory:
     Allows proper distinction between reasoning and decision taking
     No modeling restrictions (e.g. DAGs)
     It represents properly partial and total ignorance
     Ignorance is quantified:
    1. low degree of ignorance means
    – high confidence in results
    – enough information available for taking decisions
    2. high degree of ignorance means
    – low confidence in results
    – gather more information (if possible) before taking decisions
     Conflict is quantified:
    1. low conflict indicates the presence of confirming information sources
    2. high conflict indicates the presence of contradicting sources
     Simplicity: Dempster’s rule of combination covers
    1. combination of evidence
    2. Bayes’ rule
    3. Bayesian updating (conditioning)
    4. belief revision (results from non-monotonicity)

    DS-Theory is not very successful because:
     Inference is less efficient than Bayesian inference
     Pearl is the better speaker than Dempster (and Shafer, Kohlas, etc.)
     Microsoft supports Bayesian Networks
     The UAI community does not like “outsiders”

Leave an answer