Localized Partial Evaluation of Belief Networks (original) (raw)

Symbolic Probabilistic Inference in Belief Networks

National Conference on Artificial Intelligence, 1990

The Symbolic Probabilistic Inference (SPI) Algorithm (D'Ambrosio, 19891 provides an efficient framework for resolving general queries on a belief network. It applies the concept of dependency-directed backward search to probabilistic inference, and is incremental with respect to both queries and observations. Unlike most belief network algorithms, SPI is goal directed, performing only those calculations that are required to respond to

CABeN: A Collection of Algorithms for Belief Networks

1991

Belief networks have become an increasingly popular mechanism for dealing with uncertainty insystems. Unfortunately, it is known that finding the probability values of belief network nodes givena set of evidence is not tractable in general. Many different simulation algorithms for approximatingsolutions to this problem have been proposed and implemented. In this report, we describe theimplementation of a collection of such

A Best-First Search Method for Anytime Evaluation of Belief Networks

1997

We present a method for incremental evaluation of a Belief Network (BN). Evaluation is initially performed on a restricted number of nodes in the immediate vicinity of the query nodes. The BN is then traversed radially out from each query node and estimates for the belief of the query node are computed iteratively. This incremental evaluation results in a form of anytime algorithm. A best-first graph traversal strategy requires an assessment of the relative importance of various nodes in terms of contributing the most towards a query node. At each step, we must visit in priority the most significant nodes while making a trade-off with computation cost. We use the concept of arc weights in a BN to determine to what extent a node influences the query node. We also incorporate a measure of the computation cost of visiting a node, in terms of the state space sizes of the node and of its parents.