189
This paper deals with the following well known global optimization
problem of finding the global minimum of a function
Φ(
z) of
N variables
over a hyperinterval
D:
Φ(
z*) = min{Φ(
z) :
z∈
D}, (1)
D = {
z
∈
R
N
:
a
j
≤
z
j
≤
b
j
, 1<
j <
N],
(2)
where
Φ(
z) is multiextremal. If Φ(
z)
is continuous then, for the function
φ(
x) = Φ(
z(
x)),
x∈[
a,
b], (3)
where
z(
x)
is the continuous Peano-type mapping of closed interval [a, b]
onto the hyperinterval
D, we have
min{
φ(
x) :
x∈[
a,
b]} = min{Φ(
z) :
z∈
D }.
Therefore, solving the multidimensional problem (1), (2) is reduced to
solving the one-dimensional problem
φ(
x*) = min{φ(
x) :
x∈[
a,
b]}, (4)
In addition, if
Φ(
z) is Lipschitzean with a constant
K > 0 then φ(
x)
satis-
fies the Hölder condition
|
φ(
x′) – φ(
x′′)| ≤
H |
x′ –
x′′|
1/
N
,
x′,
x′′
∈[
a,
b], (5)
with a constant
H
≥ 0 (Hölder constant). For
N = 1 we will also consider the
problem (4) where the first derivative
φ(
x′) satisfies the Lipschitz condition
|
φ′(
x′) – φ′(
x′′)| ≤
L |
x′ –
x′′|,
x′,
x′′∈[
a,
b], (6)
For solving the problems (4)–(5), (4)–(6) it is proposed to use
parallel
characteristical algorithms generalizing sequential characteristical algo-
rithms. A global optimization algorithm is called a parallel characteristical
algorithm if trial points (i.e. that ones where we evaluate
φ(
x)
and for the
problem (4)–(6) also
φ′(
x))
are chosen according to the following rules.
Trials of the first
n
≥ 1 iterations are performed in arbitrary
K =
k(
n) =
p(l) +
p(2) + . . . +
p(
n)
points of the interval [
a,
b]
, where
p(
i),
i
≥ 1, denotes the number of trials of
the
i-th iteration. Trial points corresponding to any next
Q-th iteration,
Достарыңызбен бөлісу: