RUS  ENG
Full version
JOURNALS // Informatics and Automation // Archive

Tr. SPIIRAN, 2009 Issue 11, Pages 142–157 (Mi trspy52)

This article is cited in 8 papers

Synthesis of a joint graph with minimal number of edges: an algorithm formalization and a proof of correctness

V. V. Oparina, A. L. Tulupyevbc

a St. Petersburg State University of Information Technologies, Mechanics and Optics
b St. Petersburg Institute for Informatics and Automation of RAS
c St. Petersburg State University, Department of Mathematics and Mechanics

Abstract: This paper relates to machine learning of algebraic Bayesian networks (ABN).
The paper's aim is to describe an algorithm for synthesis of the secondary structure of algebraic Bayesian network as a joint graph. There is an additional constraint: the joint graph must have minimal number of edges. After the algorithm description, it’s analysis is done.
Next, we formalize input data, problem statement, and goal. Then we introduce definitions of joint graphs and weight of vertex, and later on definitions of specific elements of graph theory. Load, universal set of loads, and also the order of weights are introduced.
The last section gives a proof of the algorithm correctness. The first two statements prove that the algorithm builds a joint graph. The next two ones prove that the synthesized joint graph contains minimal number of edges.

Keywords: algorithm, algebraic Bayesian network, joint graph, universal loads set, knowledge patterns base.

UDC: 004.8



© Steklov Math. Inst. of RAS, 2024