Towards Real World Applications:
Interval-Related Talks at NAFIPS’05
On June 22–25, 2005, the 24th Annual Conference of the North American Fuzzy
Information Processing Society NAFIPS’05 was held in Ann Arbor, Michigan,
USA. Themain emphasis of this conference was on real world applications of soft
computing. In line with this emphasis, the conference was sponsored not only by
IEEE and the universities, but also by industrial sponsors, the largest of which was
the Ford Motor Company.
Forreal world applications, we need to combine different types of uncertainty.
Many different techniques have been proposed to handle different types of infor-
mation about uncertainty: Interval computations deal with interval uncertainty,
statistical techniques deal with the situations in which we know the probabilities,
fuzzy techniques deal with the situations when we only have expert knowledge
formulated in terms of words from a natural language, etc.
In many real world applications, different pieces of information come with dif-
ferent uncertainty: e.g., we may know the distribution of one of the variables and the
interval of possible values of the other variable. In such situations, it is desirable to
develop techniques that would combine different types of uncertainty. The need for
such combination was emphasized in the opening talk by Witold Perdycz, the Presi-
dent of NAFIPS, and in the two opening plenary talks by Lotﬁ A. Zadeh (University
of California, Berkeley) and by George J. Klir (Binghamton University).
Need for decision making. There already exist many techniques for combining
interval, probabilistic, and fuzzy uncertainty, and many results of using such tech-
niques have been presented at the conference. However, these techniques are mainly
oriented towards scientiﬁc applications, where the main objective is to estimate the
values of certain physical quantities. In many real world applications, it is desirable
not just to get some information about the values of physical quantities, but also to
make decisions based on this information.
Traditional approach to decision making. Traditional decision making is usually
done under probabilistic uncertainty. In this approach, for each possible action a,
and for each alternative A
n), we assume that we know the probability
(a)thattheactiona leads to the alternative A
.Wealso assume that we know the
of each alternative A
.Inthis situation, we select the action a for which
the expected value of utility u(a)
is the largest possible.
Reliable Computing 12: 73–77