Information Technology Reference
In-Depth Information
Table 3.31.
DDE Mean Flowtime
m
x
n
Generated
problems
GA
DE
(Solution)
GA/DE
4x4
5
21.38
22.11
-
5x10
5
35.3
36.34
97.14
8x15
5
63.09
66.41
95
10x25
5
98.74
103.89
95.04
15x25
5
113.85
122.59
93.03
20x50
5
216
234.32
92.18
25x75
5
317
354.77
89.35
30x100
5
399.13
435.49
91.56
Table 3.32.
EDE FSS operational values
Parameter
Values
Strategy
9
NP
150
CR
0.3
F
0.1
These obtained results formed the basic for the enhancement of DDE. It should be
noted that even with a very high percentage of in-feasible solutions obtained, DDE
managed to outperform GA.
3.6.3
Experimentation for Enhanced Differential Evolution Algorithm
The second phase of experiments outline experimentation of EDE to FSS. As with
the DDE, operational parameters were empirically obtained as given in Table 3.32. As
can be noticed the parameters are very different from those used in DDE for the same
problems. This is attributed to the new routines added to DDE which adds another layer
of stochastically to EDE.
The first section of experimentation was conducted on the same group of FSS prob-
lems as GA and DDE to obtain comparison results. In this respect, only makespan
was evaluated. For all the problem instances, EDE performs optimally compared to the
other two heuristics. Columns 5 to 7 in Table 3.33 gives the effectiveness comparisons
of EDE, DDE and GA, with EDE outperforming both DDE and GA.
With the validation completed for EDE, more extensive experimentation was con-
ducted to test its complete operational range in FSS.
The second set of benchmark problems is from the three papers of [3], [33] and [15].
All these problem sets are available in the OR Library [29]. The EDE results are com-
pared with the optimal values reported for these problems as given in Table 3.34. The
conversion is given in Equation 3.10:
Search WWH ::
Custom Search