
124 Soft Computing and Its Applications
Using the Kuhn–Tucker second-order sufciency conditions for optimality, we
can solve problem (2.104) analytically and derive the exact minimal variability OWA
weights for any level of orness.
The optimal weighting vector
W* =
is feasible if and only if , [0, 1].
We nd
,
[0, 1]
.
The following (disjunctive) partition of the unit interval (0, 1) is crucial in nding
an optimal solution to problem (2.104):
(0, 1) =
(2.105)
where
= , r = 2, …, n – 1,
= ,
= , s = 2, …, n - 1.
Consider the problem (2.104) and suppose that for some r and s from
partition (2.105). Such r and s always ...