Comparison of the performances of decision aimed algorithms with Bayesian and beliefs basis.

There are strong views in the literature between the advocates of the Bayesian approach on one hand and the advocates of the belief functions on the other. This paper refers to a few previously published studies showing that, for decision-aimed problems, algorithms using belief functions were much slower than those using a Bayesian approach. Thus, if beliefs are appealing intellectually, they are in fact useless for real applications. This article shows that most of these studies are simply false, because they are based on an erroneous use of the belief functions. © 2001 John Wiley & Sons, Inc. [ABSTRACT FROM AUTHOR]