Method and system of equivalently transforming Markov network into factor graphs

Abstract

Provided are a method and system of equivalently transforming Markov network into factor graphs, wherein the method comprises the following steps: inputting Markov network which is to be transformed into factor graphs; equivalently transforming Markov network into factor graphs, according to transformation rule 1,2,3 in sequence; and outputting the transformed factor graphs. And the transformation rule 1 is establishing all the variable nodes of factor graphs which respectively correspond to all the random factor nodes of Markov network; the transformation rule 2 is establishing all the factor nodes of factor graphs called local function, which respectively correspond to all the biggest mission style function in Markov network; and the transformation rule 3 is if and only if a certain variable in the factor graphs is a variable in the local function, undirected edges can be added to link the variable node and local function. The method and system of equivalently transforming Markov network into factor graphs have the advantages of overcoming the defects that Markov network currently lacks completeness and enforceable technical scheme in the process of equivalently transforming Markov network into factor graphs, being simple and easy to utilize and being wide in application.

Description

technical field [0001] The invention belongs to the technical field of probability graph models, and relates to a conversion technology between probability graph models, in particular to a method and system for equivalently converting a Markov network into a factor graph. Background technique [0002] Probability plays a central role in the fields of intelligent analysis and reasoning technologies such as machine learning, pattern recognition, data mining, and big data analysis. Traditional probability theory can be expressed in terms of the sum rule and the product rule equations, and all algebraic manipulations of probabilistic reasoning and learning are equivalent to repeatedly applying these two simple equations . However, with the help of the probabilistic graphical model, the underlying algebraic representations of the probabilistic model can be equivalently transformed into intuitive graphical representations, so that the complex algebraic operations of probabilistic...

AI Technical Summary

Problems solved by technology

However, there is still a lack of corresponding methods and complete and implementable technical solutions for the equivalent conversion of traditional probability graphical models to novel probability graphical models such as factor graphs.

Images