Various forms of MAs have already been reported
among a wide variety of optimization problems. Barkat
Ullah et al. (2009) proposed a new agent based on memetic
algorithm for dealing with COPs. Four types of local search
techniques are adaptively selected through learning. Liu
et al. (2007) proposed a co-evolutionary differential algo-
rithm under the concept of MAs for COPs, in which two
population are built and evolved cooperatively by inde-
pendent differential algorithm. Singh et al. (2010) pre-
sented a memetic algorithms in which the strength of
evolutionary algorithm and a local search strategy were
incorporated to tackle COPs. Gong et al. (2010) pre-
sented a hybrid algorithm based on differential evolution
and biogeography-based optimization (BBO) for global
numerical optimization problems. Kelner et al. (2008)
presented a new coupling optimization approach where a
local search strategy based on the interior point method
was integrated into genetic algorithm. More recently,
Wang et al. (2012) proposed a memetic particle swarm
optimization for tackling multi-modal optimization prob-
lems. In his work, two different local search techniques are
used in a cooperative way. Wang et al. (2009) proposed an
adaptive hill climbing strategy. The greedy crossover-
based hill climbing and steepest mutation-based hill
climbing were incorporated and used as the local search
procedure within the framework of MAs for solving
dynamic optimization problems. Tang et al. (2007) pre-
sented a diversity-based adaptive local search strategy
based on parameterized Gaussian distribution. The local
search strategy is integrated into the framework of the
parallel memetic algorithm to address large scale combi-
natorial optimization problems. Molina et al. (2010) pro-
posed a intense continuous local search in the framework
of MAs.
In the above reviews, MAs exhibit very promising
performance in various optimization problems. Therefore,
a similar hybrid idea is presented in this paper.
3 Overview of invasive weed optimization
and differential evolution
3.1 Invasive weed optimization
IWO is first presented by Mehrabian and lucas (2006)to
solve numerical optimization problems. IWO simulates the
nature principles and behaviors of weedy invasion and
colonization in the shifting and turbulent environment.
Later, Kundu et al. (2011) proposed a variant of IWO that
extends the original IWO to handle multi-objective opti-
mization problems. Generally speaking, there are four steps
of IWO.
1. Initialize a population: solutions are initialized and
dispersed in the given n dimensional search space
uniformly and randomly.
2. Reproduction: each individual of the population is
permitted to reproduce seeds according to its own
fitness, the colony’s lowest and highest fitness, in this
situation, the fitness of each individual is normalized
and the number of seeds each individual reproduces
depends on a given minimum and maximum and
increases linearly.
3. Spatial dispersal: offspring are randomly distributed
over the n dimensional search space by normally
distributed random numbers with mean equal to
zero; but varying variance. Through this, a group of
offspring are produced around their parent individual and
thus weed colony is formed to enhance the search ability.
Furthermore, standard deviation (sd) of the normally
distributed random function will be reduced from a
predefined initial value, sd
max
, toa finalvalue, sd
min
,over
every generation. the value of sd for a given generation is
computed as follows.
sd ¼
ðsd
max
sd
min
Þðiter
max
iterÞ
m
iter
m
max
þ sd
min
ð6Þ
where iter
max
is the maximum number of generations,
iter is the current number of generation and m is the
nonlinear modulation index.
4. Competitive exclusion: with the growth and reproduc-
tion of weeds, after passing several generation, the
number of weeds in a colony will reach its maximum.
Therefore, an essential exclusion mechanism is needed
among weeds. The exclusion mechanism is applied to
eliminate weeds with low fitness and select good
weeds that reproduce more than undesirable ones.
Subsequently, the selected ones will be preserved into
the next generation and then the steps I-IV are repeated
until termination criterion is reached.
3.2 Differential evolution
DE, proposed by Storn and Price (1995), is a powerful
search algorithm in the optimal problems, and uses the
vector differences of individuals for perturbing the popu-
lation members.
Initially, DE comprises a population of N and every
individual is an n-dimensional vectors x
i
¼fx
1
; x
2
; ...;
x
n
g: These vectors are randomly generated in the search
space and in the process of evolution, individuals will be
tackled by the operations of mutation, crossover and
selection.
Mutation operation: in this operation, with the different
mutant strategies, the generated way of a mutant vector v
i
1896 X. Cai et al.
123