Information Technology Reference
In-Depth Information
Fig. 4.
The final
-nondominated solutions for ZDT1 problem, obtained by (a) omni-
aiNet (79 solutions) and (b) DT omni-optimizer.
Fig. 5.
The final
-nondominated solutions for ZDT2 problem, obtained by (a) omni-
aiNet (92 solutions) and (b) DT omni-optimizer
6.4
Multi-objective Multi-global Problem
The multi-objective multi-global problem used in this work was designed by Deb
and Tiwari in [8] and is given by:
Minimize
f
1
(
x
)=
i
=1
sin(
πx
i
)
,
f
2
(
x
)=
i
=1
cos(
πx
i
)
,
(6)
subject to
0
≤
x
i
≤
6
Both objectives of this problem are periodic functions with period 2, such
that Pareto optimal solutions correspond to
x
i
[2
m
+1
,
2
m
+3
/
2], where
m
is an integer. Figure 6 presents the
-nondominated solutions obtained by
(a) omni-aiNet and (b) DT omni-optimizer. The omni-aiNet parameters were
100 generations, 400 individuals in the initial population, 3 generations between
suppressions, 200 individuals in Random Insertion, a suppression threshold of
0
.
006, 2 clones per individual and
δ
=0
.
001. For this problem, the obtained
results for DT omni-optimizer were achieved with
δ
=0
.
001, 100 generations
and a population of 400 individuals.
∈