TY - GEN

T1 - Statistical Comparison of the Running Times of Global Posteriori Inference Algorithms in Algebraic Bayesian Networks

AU - Вяткин, Артём Андреевич

AU - Тулупьев, Александр Львович

PY - 2023/5/24

Y1 - 2023/5/24

N2 - Global posteriori inference is an important problem in the theory of algebraic Bayesian networks, which are a subclass of probabilistic graphical models. So far, two algorithms have been presented that allow global posteriori inference. These algorithms use different global algebraic Bayesian network structures – secondary and tertiary. Thus, the task of comparing these algorithms is defined, and, in particular, it is necessary to perform a statistical analysis of their running times, which is the purpose of this paper. In addition, the stochastic generation algorithm of acyclic algebraic Bayesian networks was defined in the framework of the article, which allowed to generate data for the experiment. As a result, it was found that the algorithm using a tertiary structure works 1.15–2 times faster on algebraic Bayesian networks, where there are 5–150 knowledge patterns. At the same time, the difference in running time is noticeable the more atoms in the knowledge pattern models and their intersections.

AB - Global posteriori inference is an important problem in the theory of algebraic Bayesian networks, which are a subclass of probabilistic graphical models. So far, two algorithms have been presented that allow global posteriori inference. These algorithms use different global algebraic Bayesian network structures – secondary and tertiary. Thus, the task of comparing these algorithms is defined, and, in particular, it is necessary to perform a statistical analysis of their running times, which is the purpose of this paper. In addition, the stochastic generation algorithm of acyclic algebraic Bayesian networks was defined in the framework of the article, which allowed to generate data for the experiment. As a result, it was found that the algorithm using a tertiary structure works 1.15–2 times faster on algebraic Bayesian networks, where there are 5–150 knowledge patterns. At the same time, the difference in running time is noticeable the more atoms in the knowledge pattern models and their intersections.

KW - knowledge pattern

KW - probabilistic-logic inference

KW - tertiary structure

KW - probabilistic graphical models

KW - machine learning

KW - statistical study of algorithm complexities

UR - https://www.mendeley.com/catalogue/0fa1f66b-d855-3d4f-b5e2-f83d4532feb2/

U2 - 10.1109/scm58628.2023.10159044

DO - 10.1109/scm58628.2023.10159044

M3 - Conference contribution

SN - 9798350322484

SP - 24

EP - 28

BT - Proceedings of 2023 XXVI International Conference on Soft Computing and Measurements (SCM)

T2 - 2023 XXVI International Conference on Soft Computing and Measurements (SCM)

Y2 - 24 May 2023 through 26 May 2023

ER -