## Stable Local Computation with Conditional Gaussian Distributions (1999)

Venue: | Statistics and Computing |

Citations: | 68 - 2 self |

### BibTeX

@ARTICLE{Lauritzen99stablelocal,

author = {Steffen L. Lauritzen and Frank Jensen},

title = {Stable Local Computation with Conditional Gaussian Distributions},

journal = {Statistics and Computing},

year = {1999},

volume = {11},

pages = {191--203}

}

### Years of Citing Articles

### OpenURL

### Abstract

: This article describes a propagation scheme for Bayesian networks with conditional Gaussian distributions that does not have the numerical weaknesses of the scheme derived in Lauritzen (1992). The propagation architecture is that of Lauritzen and Spiegelhalter (1988). In addition to the means and variances provided by the previous algorithm, the new propagation scheme yields full local marginal distributions. The new scheme also handles linear deterministic relationships between continuous variables in the network specification. The new propagation scheme is in many ways faster and simpler than previous schemes and the method has been implemented in the most recent version of the HUGIN software. Key words: Artificial intelligence, Bayesian networks, CG distributions, Gaussian mixtures, probabilistic expert systems, propagation of evidence. 1 Introduction Bayesian networks have developed into an important tool for building systems for decision support in environments characterized by...

### Citations

7319 |
Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference
- Pearl
- 1988
(Show Context)
Citation Context ...expert systems, propagation of evidence. 1 Introduction Bayesian networks have developed into an important tool for building systems for decision support in environments characterized by uncertainty (=-=Pearl 1988-=-; Jensen 1996; Cowell et al. 1999). The exact computational algorithms that are most developed are concerned with networks involving discrete variables only. This is Research Report R-99-2014, Departm... |

1322 | Local computations with probabilities on graphical structures and their application to expert systems (with discussion - Lanritzen, Spiegelhalter - 1988 |

1132 | Graphical models
- Lauritzen
- 1996
(Show Context)
Citation Context ... continuous variables depends linearly on the continuous variables for fixed values of the discrete variables (i; j). In a general CG regression, p is also permitted to depend on z in a specific way (=-=Lauritzen 1996-=-), but this is not relevant here. 3 Mixed Bayesian networks We consider probabilistic networks over a directed acyclic graph (DAG), known as Bayesian networks (Pearl 1986). A mixed Bayesian network wi... |

950 |
An Introduction to Bayesian Networks
- Jensen
- 1998
(Show Context)
Citation Context ...ms, propagation of evidence. 1 Introduction Bayesian networks have developed into an important tool for building systems for decision support in environments characterized by uncertainty (Pearl 1988; =-=Jensen 1996-=-; Cowell et al. 1999). The exact computational algorithms that are most developed are concerned with networks involving discrete variables only. This is Research Report R-99-2014, Department of Mathem... |

649 |
Probabilistic Networks and Expert Systems
- Cowell, Dawid, et al.
- 1999
(Show Context)
Citation Context ...on of evidence. 1 Introduction Bayesian networks have developed into an important tool for building systems for decision support in environments characterized by uncertainty (Pearl 1988; Jensen 1996; =-=Cowell et al. 1999-=-). The exact computational algorithms that are most developed are concerned with networks involving discrete variables only. This is Research Report R-99-2014, Department of Mathematical Sciences, Aal... |

387 | Fusion, Propagation, and Structuring in Belief Networks
- Pearl
- 1986
(Show Context)
Citation Context ...d on z in a specific way (Lauritzen 1996), but this is not relevant here. 3 Mixed Bayesian networks We consider probabilistic networks over a directed acyclic graph (DAG), known as Bayesian networks (=-=Pearl 1986-=-). A mixed Bayesian network with conditional Gaussian distributions is specified over a set of nodes or variablessV , partitioned as V = \Delta [ \Gamma into discrete and continuous variables as above... |

290 |
Bayesian updating in causal probabilistic networks by local computations
- Jensen, Lauritzen, et al.
- 1990
(Show Context)
Citation Context ...the initialization process involves sending messages from the leaves of the junction tree towards the root in a way similar to the process known as CollectEvidence in the standard HUGIN architecture (=-=Jensen et al. 1990-=-), although the messages sent are slightly different. Thus, a clique is allowed to send a message if it is a leaf of the junction tree, or if it has received messages from all of its neighbours furthe... |

210 |
A generalised inverse for matrices
- Penrose
- 1955
(Show Context)
Citation Context ... \Gamma C 21 C \Gamma 11 A 1 F = [ C 21 C \Gamma 11 : B 2 \Gamma C 21 C \Gamma 11 B 1 ] G = C 22 \Gamma C 21 C \Gamma 11 C 12 : Here M \Gamma denotes an arbitrary generalized inverse of the matrix M (=-=Penrose 1955-=-), i.e. an arbitrary matrix M \Gamma satisfying MM \Gamma M = M; (2) see also Rao (1973), pp. 24--27, and Rao and Mitra (1971). Then [p; A; B; C](H j T ) = [p ; A 1 ; B 1 ; C 11 ](H 1 j T ) \Omega [q;... |

174 | Graphical Models for Associations between Variables, some of which are Qualitative and some Quantitative,” The - Lauritzen, Wermuth - 1989 |

143 | Propagation of probabilities, means, and variances in mixed graphical association models - Lauritzen - 1992 |

61 | Probabilistic Expert Systems - Shafer - 1996 |

53 | Linear Statistic Inference - Rao - 1973 |

37 | Lazy propagation in junction trees
- Madsen, Jensen
- 1998
(Show Context)
Citation Context ...ion of head variables given their tail. But as the partitioning of discrete variables is not exploited in our propagation scheme, we have chosen not to do so. A propagation scheme of the `lazy' type (=-=Madsen and Jensen 1998-=-) could exploit such a partitioning. The initial conditional distribution for a continuous variable v with parent nodes pa(v) in a mixed Bayesian network corresponds to the CG potentials[1; ff; fi ? ;... |

28 | Local computation with valuations from commutative semigroups - Lauritzen, Jensen - 1997 |

19 | An axiomatic study of computation in hypertrees - Shafer - 1991 |

18 | Mixed Interaction Models - Lauritzen, Wermuth - 1984 |