Structured probabilistic inference
Under an Elsevier user license
open archive
Highlights
► This paper presents new inference algorithms in Object-Oriented Bayesian Networks, based on Probabilistic Relational Models. ► After a survey, we present the version of PRMs used in this work. ► We then adapt state-of-the-art Structured Variable Elimination (SVE) algorithm to our framework. ► We show that some drawbacks of SVE can be removed in a new algorithm called SPI. ► In order to speed-up the inference, we include a d-separation analysis at class level, leading to a new algorithm SPISBB which outperforms other algorithms.
Keywords
Bayesian Network
Probabilistic Graphical Models
Probabilistic Relational Models
Object Oriented Bayesian Networks
Probabilistic inference
Structured inference
Cited by (0)
Copyright © 2012 Published by Elsevier Inc.