A new method is proposed for exploiting causal independencies in exact Bayesian network inference. A Bayesian network can be viewed as representing a factorization of a joint probability into the multiplication of a set of conditional probabilities. We present a notion of causal independence that enables one to further factorize the conditional probabilities into a combination of even smaller factors and consequently obtain a finer-grain factorization of the joint probability. The new formulation of causal independence lets us specify the conditional probability of a variable given its parents in terms of an associative and commutative operator, such as “or”, “sum ” or “max”, on the contribution of each parent. We start with a simple algorithm VE for Bayesian network inference that, given evidence and a query variable, uses the factorization to find the posterior distribution of the query. We show how this algorithm can be extended to exploit causal independence. Empirical studies, based on the CPCS networks for medical diagnosis, show that this method is more efficient than previous methods and allows for inference in larger networks than previous algorithms. 1.
|
4388
|
Probabilistic Reasoning in Intelligent Systems
– Pearl
- 1988
|
|
904
|
Local computations with probabilities on graphical structures and their applications to expert systems
– Lauritzen, Spiegelhalter
- 1988
|
|
411
|
The computational complexity of probabilistic inference using Bayesian belief networks
– Cooper
- 1990
|
|
235
|
uence diagrams
– Howard, Matheson
- 1981
|
|
231
|
Complexity of finding embeddings in a k-tree
– Arnborg, Corneil, et al.
- 1987
|
|
220
|
Bayesian updating in causal probabilistic networks by local computations
– Jensen, Lauritzen, et al.
- 1990
|
|
183
|
Bucket elimination: A unifying framework for probabilistic inference
– Dechter
- 1999
|
|
176
|
Context-specific independence in bayesian networks
– Boutilier, Friedman, et al.
- 1996
|
|
150
|
Nonserial Dynamic Programming
– Bertele, Brioschi
- 1972
|
|
112
|
Independence properties of directed Markov fields
– Lauritzen, Dawid, et al.
- 1990
|
|
91
|
Knowledge engineering for large belief networks
– Pradhan, Provan, et al.
- 1994
|
|
90
|
Subjective bayesian methods for rulebased inference systems
– Duda, Hart, et al.
- 1976
|
|
78
|
Probability propagation
– Shafer, Shenoy
- 1990
|
|
75
|
Some Practical Issues in Constructing Belief Networks
– Henrion
- 1989
|
|
71
|
A model of inexact reasoning in medicine
– Shortliffe, Buchanan
- 1975
|
|
65
|
Knowledge representation and inference in similarity netwrosk and Bayesian multinets
– Geiger, Heckerman
- 1996
|
|
55
|
A computational model for causal and diagnostic reasoning in inference engines
– Kim
- 1983
|
|
54
|
A new look at causal independence
– Heckerman, Breese
- 1994
|
|
52
|
A generalization of the noisy-or model
– Srinivas
- 1993
|
|
49
|
Simple linear time algorithms to test chordiality of graphs, test acyclicity of graphs, and selectively reduce acyclic hypergraphs
– Tarjan, Yannakakis
|
|
45
|
Symbolic probabilistic inference in belief networks," Automated Reasoning
– Shachter, D'Ambrosio, et al.
- 1990
|
|
41
|
Parameter adjustment in Bayes networks. The generalized noisy OR-gate
– Díez
- 1993
|
|
32
|
Causal independence for knowledge acqui-sition and inference
– Heckerman
- 1993
|
|
32
|
E cient Inference in Bayes Networks as a Combinatorial Optimization Problem
– Li, D'Ambrosio
- 1994
|
|
30
|
d�separation� From theorems to algorithms
– Verma�, Pearl
- 1989
|
|
25
|
Local expression languages for probabilistic dependence
– D'Ambrosio
- 1991
|
|
25
|
A Munin network for the median nerve - a case study on loops
– Olesen, Kjaerulff, et al.
- 1989
|
|
19
|
A causal calculus
– Good
- 1961
|
|
17
|
Pruning Bayesian networks for efficient computation
– Baker, Boult
- 1990
|
|
16
|
Using causal knowledge to create simulated patient cases: the CPCS project as an extension of INTERNIST-1
– Parker, Miller
- 1987
|
|
15
|
Triangulation of graphs - algorithms giving small total state space
– Kjærulff
- 1990
|
|
9
|
Symbolic probabilistic inference in large BN2O networks
– D'Ambrosio
- 1994
|
|
9
|
Parameter adjustment in Bayes networks: the generalized noisy OR-gate
– Dez
- 1993
|
|
6
|
Additive belief-network models
– Dagum, Galper
- 1993
|
|
5
|
Triangulation of graphs - algorithms giving small total state space
– Kj��rulff
- 1990
|
|
4
|
Specification of models in large expert systems based on causal probabilistic networks
– Olesen, Andreassen
- 1993
|
|
1
|
The computational complexity of probabilisticinference using Bayesian belief networks
– Cooper
- 1990
|
|
1
|
Additivebelief-network models
– Dagum
- 1993
|
|
1
|
Knowledgerepresentation and inference in similarity networks and Bayesian multinets
– Geiger
- 1996
|
|
1
|
ProbabilisticHorn abduction and Bayesian networks
– Poole
- 1993
|
|
1
|
Probabilistic Horn abductionand Bayesian networks
– Poole
- 1993
|
