Information Technology Reference
In-Depth Information
Table 2.
Comparimion of
N
max Value of Different Size
N
max
15
30
50
75
100
Ta001(20x5)
0
0
0
2.32
1.41
Car1(11x5)
0
0
0
0
0
Car3(12x5)
1.79
1.5
1.79
3.0
1.79
Car7(7x7)
2.87
0.8
0.8
0.8
0.8
reC03(20x5)
2.06
1.97
2.51
2.43
1.98
Ta001(20x5)
0
0
0
2.32
1.41
Then, the frog memeplex is fixed to 10, and the local exploration is set to 20. The
number of frogs in each sub-memeplex is set to 12, 25, 33, 50, and 66. Fig. 4 shows
the variance of MDP with frogs' amount in sub-memeplexes. At last, the number of
frogs in sub-memeplexes is fixed to 30, the number of local iteration for each sub-
memeplex is set to 20, and the number of sub-memeplex is set to 5, 10, 15, 20, and
30. We can get the comparison of different size problems in the test.
Fig. 4.
Comparison of the different number of frogs in sub-memeplexes
Based on those experimental results, the suitable values are found as follows: the
number of memeplexes and the number of frogs in each sub-memeplex are 15 and 25,
respectively, number of processing cycles on each memeplex before shuffling is 30.
To further analyze the performance of the proposed method ISFLA, we compared
our proposed algorithm on the same type of scheduling problems with the general
SFLA. In the simulation, we used some benchmarks which would be fuzzed as the
test examples. The triangular number
~
(
)
is applied to denote the
O
L
,
M
,
O
U
processing time of experimental job, where
O
M
is data of some typical hard
benchmarks,
O
and
O
can be got by
.
O
M
−
rand
(
0
,
M
*
0
.
2
)
,
O
M
+
rand
(
0
,
M
*
0
.
2
)
The due-date window is assumed by
.
The earliness weights
h
and tardiness weights
w
are uniformly generated in the
interval [1, 6], and each instance was randomly performed 10 times. Statistic results
are shown in Table 3 and Table 4.
[
ms
−
rand
(
0
,
ms
*
0
.
2
)
,
ms
+
rand
(
0
,
ms
*
0
.
2
)]
Search WWH ::
Custom Search