PII: 0004-3702(96)80446-2


Artificial 

ELSEVIER Artificial Intelligence 82 (1996) 393-394 

Intelligence 

Forthcoming Papers 

A. Becker and D. Geiger, Optimization of Pearl’s method of conditioning and 
greedy-like approximation algorithms for the vertex feedback set problem 

We show how to find a small loop curser in a Bayesian network. Finding such a loop cutset is the first step in 
the method of conditioning for inference. Our algorithm for finding a loop cutset, called MGA, finds a loop 
cutset which is guaranteed in the worst case to contain less than twice the number of variables contained in 
a minimum loop cutset. The algorithm is based on a reduction to the weighted vertex feedback set problem 
and a new approximation of the latter problem. The complexity of MGA is O(NI + n logn) where m and n 
are the number of edges and vertices respectively. A greedy algorithm, called GA, for the weighted vertex 
feedback is also analyzed and bounds on its performance are given. We test MGA on randomly generated 
graphs and find that the average ratio between tbe number of instances associated with the algorithm’s output 
and the number of instances associated with an optimum solution is 1.22 for the graphs tested. 

R Walley, Measures of uncertainty in expert systems 

This paper compares four measures that have been advocated as models for uncertainty in expert systems. The 
measures are additive probabilities (used in the Bayesian theory), coherent lower (or upper) previsions, belief 
functions (used in the Dempster-Shafer theory) and possibility measures (fuzzy logic). Special emphasis is 
given to the theory of coherent lower previsions, in which upper and lower probabilities, expectations and 
conditional probabilities are constructed from initial assessments through a technique of natural extension. 
Mathematically, all the measures can be regarded as types of coherent lower or upper previsions, and this per- 
spective gives some insight into the properties of belief functions and possibility measures. The measures are 
evaluated according to six criteria: clarity of interpretation; ability to model partial information and imprecise 
assessments, especially judgements expressed in natural language; rules for combining and updating uncer- 
tainty, and their justification; consistency of models and inferences; feasibility of assessment; and feasibility 
of computations. Each of the four measures seems to be useful in special kinds of problems, but only lower 
and upper previsions appear to be sufficiently general to model the most common types of uncertainty. 

Y. Moses and M. Tennenholtz, Off-line reasoning for on-line efficiency: knowledge 
bases 

The complexity of reasoning is a fundamental issue in AI. In many cases, the fact that an intelligent system 
needs to perform reasoning on-line contributes to the difficulty of this reasoning. This paper considers the 
case in which an intelligent system computes whether a query is entailed by the system’s knowledge base. 
It investigates how an initial phase of off-line preprocessing and design can improve the on-line complexity 

Elsevier Science B.V.