Information Technology Reference
In-Depth Information
n
1
n
M
(
x
1
,...,
x
n
)
=
x
i
,
i
=
1
1
n
and with
f
(
x
)
=−
log
x
, and
p
1
=
(
1
i
n
)
is the geometric mean
n
√
p
1
M
(
x
1
,...,
x
n
)
=
·
p
2
···
p
n
.
x
1
x
ʱ
(ʱ >
,is
f
−
1
1
With
f
(
x
)
=
0
)
(
x
)
=
ʱ
, and with
p
1
=
n
, it is obtained the
family of quasi-linear means,
x
1
1
ʱ
x
n
+ ··· +
M
ʱ
(
x
1
,...,
x
n
)
=
.
n
In particular, with
ʱ
=
1, is
M
1
the arithmetic mean, and with
ʱ
=−
1, it follows
n
M
(
x
1
,...,
x
n
)
=
(
provided
x
1
,...,
x
n
=
0
),
−
1
1
1
x
n
x
1
+ ··· +
called
Harmonic Mean
. As it is easy to prove,
n
√
x
1
···
lim
ʱ
ₒ
M
ʱ
(
x
1
,...,
x
n
)
=
x
n
0
lim
ʱ
ₒ∞
M
ʱ
(
x
1
,...,
x
n
)
=
max
(
x
1
,...,
x
n
)
lim
ʱ
ₒ−∞
M
ʱ
(
x
1
,...,
x
n
)
=
min
(
x
1
,...,
x
n
).
2.3.2 Ordered Weighted Means
n
It is said that
M
:[
0
,
1
]
ₒ[
0
,
1
]
is a
mean
, when
M
is continuous, monotonic,
and verifies:
min
(
x
1
,...,
x
n
)
M
(
x
1
,...,
x
n
)
max
(
x
1
,...,
x
n
).
Since, min
(
0
,...,
0
)
M
(
0
,...,
0
)
max
(
0
,...,
0
)
=
0, it results
M
(
0
,...,
0
)
=
0. Since, min
(
1
,...,
1
)
M
(
1
,...,
1
)
max
(
1
,...,
1
)
=
1, it results
1. Hence, of course, quasi-linear means are means, but there are
more of such means. An important and useful example are the
Ordered Weighted
Means
(OWA). Its definition is the following:
M
(
1
,...,
1
)
=
Search WWH ::
Custom Search