Abstract de nuestro grupo en 1995/UTAI Research Group Abstracts in 1995


Heuristic Algorithms for the Triangulation of Graphs
BY Andrés Cano and Serafín Moral

To appear in Advances in Intelligent Computing. ed: B. Bouchon-Meunier, R.R. Yager and L.A. Zadeh. Springer Verlag, 1995. (10 pages)

Different uncertainty propagation algorithms in graphical structures can be viewed as a particular case of propagation in a joint tree, which can be obtained from different triangulations of the original graph. The complexity of the resulting propagation algorithms depends on the size of the resulting triangulated graph. The problem of obtaining an optimum graph triangulation is known to be NP-complete. Thus approximate algorithms which find a good triangulation in reasonable time are of particular interest. This work describes and compares several heuristic algorithms developed for this purpose.


Independence Concepts for Convex Sets of Probabilities
BY Luis M. de Campos and Serafín Moral

To appear in: Proceedings of the Eleventh Conference on Uncertainty in Artificial Intelligence (Ph. Besnard, S. Hanks, eds.) Morgan Kaufmann (San Mateo), 1995, (8 pages).

In this paper we study different concepts of independence for convex sets of probabilities. There will be two basic ideas for independence. The first is irrelevance. Two variables are independent when a change on the knowledge about one variable does not affect the other. The second one is factorization. Two variables are independent when the joint convex set of probabilities can be decomposed on the product of marginal convex sets. In the case of the Theory of Probability, these two starting points give rise to the same definition. In the case of convex sets of probabilities, the resulting concepts will be strongly related, but they will not be equivalent. As application of the concept of independence, we shall consider the problem of building a global convex set from marginal convex sets of probabilities.


Revision rules for convex sets of probabilities
BY S. Moral and N. Wilson

To appear in: Mathematical Models for Handling Partial Knowledge in Artificial Intelligence (G. Coletti, D. Dubois, R. Scozzafava, eds.) Plenum Press, 1995.

This paper investigates the use of a class of importace sampling algorithms for probabilistic graphs in graphical structures. A general model for constructing importance sampling algorithms is given and then some particular cases are considered. Logical sampling and likelihood weigthing are particular cases of the model. Our proposal will be an algorithm which uses the functions with less entropy (more informative) to simulate the variables and the functions with more entropy to weight the simulations, in this way we expect to obtain more uniform weights. Some experimental tests are carried out comparing the performance of the proposed algorithms in randomly generated graphs.


Axiomatic Treatment of Possibilistic Independence
BY Luis M. de Campos, Joerg Gebhardt and Rudolf Kruse

In Symbolic and Quantitative Approaches to Reasoning and Uncertainty, Lecture Notes in Artificial Intelligence 946, C. Froidevaux and J. Kohlas (Eds.), 77-88, Springer Verlag (1995).

The clarification of the concepts of independence, marginalization and combination of modularized information is one of the major topics concerning the efficient treatment of imperfect data in complex domains of knowledge. Confining to the uncertainty calculus of possibility theory, we consider a syntactic (based on a set of axioms) as well as a semantic approach (in a random set framework) to appropriate definitions of possibilistic independence. It turns out that well-known, but also new proposals for the concept of possibilistic independence can be justified.


Propagación de probabilidades en grafos de dependencias mediante muestreo por importancia
BY J.E. Cano, L.D. Hernández, S. Moral

Actas de la VI Conferencia de la Asociación Española para la Inteligencia Artificial (CAEPIA 95) 197-206., 1995. (10 pages)

Los esquemas de simulación para la propagación de probabilidades en grafos de dependencias presentan ciertas ventajas frente a los usuales métodos exactos. Un grupo de tales esquemas reciben el nombre de métodos basados en Muestreo Hacia Adelante. En este trabajo presentamos una clase general de algoritmos que se engloba en este grupo, basados en la conocida técnica del muestreo por importancia. Esta clase incluye a los conocidos métodos del muestreo lógico y de la ponderación de la verosimilitud como casos particulares, y elimina los inconvenientes que presentan ambos métodos introduciendo criterios de entropía. Se lleva a cabo una experimentación comparando su comportamiento con el de la ponderación de la verosimilitud.



Ir Atrás         Página DECSAI
Go back          Home DECSAI