Untitled in english

Author(s)

Jaime Shinsuke Ide

• Document type: Master's Dissertation
• Press: São Paulo.
• Institution: Universidade de São Paulo (USP). Escola Politécnica (EP/BC)
• Defense date: 2002-11-22

Examining board members

Fabio Gagliardi Cozman; Marcia D Elia Branco; Gilberto Francisco Martha de Souza

• Advisor: Fabio Gagliardi Cozman
• Field of knowledge: Physical Sciences and Mathematics - Computer Science
• Indexed in: Banco de Dados Bibliográficos da USP-DEDALUS
• Location: Universidade de São Paulo. Biblioteca Central da Escola; EPBC/FD-3288

Abstract

Bayesian networks are employed in Artificial Intelligence to represent uncertainty. No algorithm in the literature currently offers guarantees concerning the distribution of generated Bayesian networks. This work presents new methods for random generation of Bayesian networks. Such methods can be used to test inference and learning algorithms for Bayesian networks, and to obtain insights on average properties of such networks. This work proposes new algorithms that can generate uniformly distributed samples of directed a cyclic graphs, like multi-connected networks and polytrees, for a given number of nodes and arcs. After a directed a cyclic graph is uniformly generated, the conditional distributions are produced by sampling Dirichlet distributions. The main result of this work is the development of a freely distributed random Bayesian network generator, BNGenerator. An application of random generated Bayesian networks in the analysis of quasi-Monte Carlo methods is presented. (AU)