Intelligent Systems Laboratory
Reasoning
Reasoning under uncertainty is one component in the evaluation
function of a state-space search
which
can be used both by
procedures for automatically determining a vision algorithm
and procedures in the search part of a vision
algorithm operating on a particular image.
A variety of techniques exist for
modeling this reasoning process. There are
Bayesian and fuzzy approaches for handling the probability of a statement
being valid and there are belief approaches for handling beliefs
for combined evidence. The Bayesian approaches have severe
limitations in the nature of the dependencies they permit and the
approaches for combining evidence using belief functions often do
not have an operational meaning. In this project we are developing an
approach which is simultaneously Bayesian in character without the
usual dependency restrictions and yet a
generalization of the Shafer and Dempster approach for evidential reasoning.
