文档介绍：A Survey of Algorithms for Real-Time work Inference
Haipeng Guo William Hsu
Laboratory for Knowledge Discovery in Databases
Department puting and Information Sciences, Kansas State University
******@ ******@
Abstract
2 PRELIMINARIES
As works are applied to more
complex and realistic real-world applications, the works
development of more efficient inference BNs (also known as Bayesian works, causal
algorithms working under real-time constraints is
networks, or works) are currently the
ing more and more important. This paper
dominant uncertainty knowledge representation and
presents a survey of various exact and
approximate work inference reasoning technique in AI [Pe88, Ne90, RN95, CDLS99].
algorithms. In particular, previous research on BNs are directed acyclic graphs (DAGs) where nodes
real-time inference is reviewed. It provides a represent random variables, and edges represent
framework for understanding these algorithms conditional dependencies between random variables.
and the relationships between them. Some These random variables can be either continuous or
important issues in real-time works discrete. For simplicity, in this paper we shall only
inference are also discussed. consider discrete ones.
Definition – work: A work is a
1 INTRODUCTION graph in which the following holds [RN95]:
Over the last 20 years or so, works (BNs)
[Pe88, Ne90, RN95, CDLS99] have e the key ÿ
A set of random variables makes up the nodes of the
method for representation and reasoning work.
uncertainty in AI. BNs not only provide a natural and
ÿ
A set of directed links connects pairs of nodes. The
compact way to encode exponentially sized joint
intuitive meaning of an arrow from node X to node Y
probability distributions, but also provide a basis for
is that X has a direct influence on Y.
efficient probabilistic inference. Although there exists
polynomial time inference algorithm for specific classes