Databases Reference
In-Depth Information
SPEA2 maintains two types of population,
P
h
(called the “regular population,”
size
j
P
j
)and
A
h
(called the “archive,” size
j
A
j
), for each generation h, which are
updated, basically, as follows.
1. First generation (h
D
0):
P
0
is the initial population of
j
P
j
individuals (result
of applying METIS with
j
P
j
different seeds) and
A
0
is set to empty.
2. Generation .h
C
1/:
a.
A
h
C
1
D
set of non-dominated individuals of
P
h
[
A
h
(truncated to
j
A
j
individual if the population size is larger than
j
j
, or padded with lowest-
fitness individuals among the dominated if the population size is less than
j
A
A
j
)
b.
P
h
C
1
D
set of
j
P
j
individuals after application of recombination and
A
h
C
1
When the maximum number of generations, h
, is reached, the final partition
assignments will be the non-dominated individuals in the final archive
mutation on
A
h
.An
individual
p
dominates an individual
q
, denoted by
p q
if the corresponding
partition assignment of
p
is no worse than that of
q
in terms of both L
write
,
with at least one objective
strictly
better. The fitness of an individual
p
in generation
h is defined as
read
and G
X
1
p
;k
C
2
fitness.
p
/
D
strength.
q
/
C
q
2
P
h
[
A
h
W
q
p
where
strength
.
q
/ is the number of individuals dominated by
p
in the set
A
h
and
p
;k
denotes the distance in the objective
space from
individual
p
to its k-nearest
P
h
[
individual. As common setting, k is set to
p
j
P
jCj
A
j
.
3.3
Numerical Results
We present here the results of a simulation study to evaluate S-PUT with SPEA2,
using the Facebook dataset made available by Max-Planck Software Institute for
Software Systems [
39
]. This dataset contains a Facebook network of N
D
63; 392
users in New Orleans region, with 816,886 links resulting an average degree of 25.7.
We assume that the read and write rates of a user are proportional to its social degree,
thus these rates are given values in the range (0, 1) proportional to the degree. The
social strength between two neighboring nodes is uniformly generated, also in the
range (0, 1).
The number of servers is set to
M
D
16. The parameters for SPEA2 are:
regular population size
j
j
, recombination
probability p
recombination
D
0:8, mutation probability p
mutation
D
0:5, gene mutation
probability p
gene mutation
D
0:001, and number of generations h
D
100.
P
j2f
100; 500
g
,archivesize
j
A
jDj
P
Search WWH ::
Custom Search