An approach to the sensitivity analysis of local aposterior inference equations in algebraic Bayesian networks is proposed in the paper. Basic definitions and formulations are briefly given and the development of the matrix-vector approach of a posterior inference is considered. The propagation of deterministic and stochastic evidences in a knowledge pattern with scalar estimates of probabilities of truth over quantum propositions is described. For each of the provided cases necessary metrics are introduced and transformations which result into construction of 4 linear programming problems which solution gives the required estimates are performed. In addition, 2 theorems that postulate the covering estimates for both cases are formulated. The results obtained in the paper prove the correctness of models and create a basis for investigation of local and global logic-probabilistic inference equations sensitivity.
Translated title of the contributionSensitivity statistical estimates of local aposterior inference matrix-vector equations in algebraic Bayesian networks on quanta propositions
Original languageRussian
Pages (from-to)60-69
JournalВЕСТНИК САНКТ-ПЕТЕРБУРГСКОГО УНИВЕРСИТЕТА. МАТЕМАТИКА. МЕХАНИКА. АСТРОНОМИЯ
Volume63
Issue number1
StatePublished - 2018

    Research areas

  • uncertain knowledge, evidence propagation, probabilistic logic, algebraic Bayesian networks, probabilistic-logic inference, sensitivity statistical estimate, probabilistic graphical model, matrix-vector equations

ID: 36985955