Andr�s Cano, Manuel G�mez, Andr�s R. Masegosa, Seraf��n Moral

---

Comparing Binary and Standard Probability Trees in Credal Networks Inference

Abstract

This paper proposes the use of Binary Probability Trees in the propagation of credal networks. Standard and binary probability trees are suitable data structures for representing potentials because they allow to control the accuracy of inference algorithms by means of a threshold parameter. The choice of this threshold is a trade-off between accuracy and computing time. Binary trees enable the representation of finer-grained independences than probability trees. This leads to more efficient algorithms for credal networks with variables with more than two states. The paper shows experiments comparing binary and standard probability trees in order to demonstrate their performance.

Keywords

Credal networks, Inference algorithms, Imprecise probabilities, Bayesian networks, Variable elimination, Probability trees

---

Download area

The paper is available in the following formats:

• Gzipped pdf file
• pdf file

Plenary talk: file

Poster: file

---

Authors’ addresses

Andr�s Cano
Dpto. Ciencias de la Computaci�n e I.A.
ETS Ingenier�as Inform�tica y de Telecomunicaci�n
Avda. Andalucia s/n
Granada 18071
CIF: Q1818002F
Spain

Manuel G�mez
Dpto. Ciencias de la Computaci�n e I.A.
ETS Ingenier�as Inform�tica y de Telecomunicaci�n
C// Periodista Daniel Saucedo Aranda
18071 Granada

Andr�s R. Masegosa
Departamento de Ciencias de la Computaci�n e Inteligencia Artificial
Escuela T�cnica Superior de Ingenier�as Inform�tica y de Telecomunicaci�n
c/. Daniel Saucedo Aranda, s/n 18071 Granada

Seraf��n Moral
Dpto. Ciencias de la Computaci�n e IA
ETSI Inform�tica
Universidad de Granada
18071 Granada
SPAIN

E-mail addresses

Andr�s Cano		acu@decsai.ugr.es
Manuel G�mez		mgomez@decsai.ugr.es
Andr�s R. Masegosa		andrew@decsai.ugr.es
Seraf��n Moral		smc@decsai.ugr.es