International Journal of Communications and Information Technology, IJCIT IJCIT-2011-Vol.1-No.1 Dec.

2011

Improved Cuckoo Search Algorithm for

Global Optimization
Ehsan Valian, Shahram Mohanna and Saeed Tavakoli
University of Sistan and Baluchestan

The Cuckoo Search (CS) developed by Yang and Deb

(Yang 2009; Yang 2010), is a new meta-heuristic algorithm
Abstract- The cuckoo search algorithm is a recently developed imitating animal behavior. The optimal solutions obtained
meta-heuristic optimization algorithm, which is suitable for by the CS are far better than the best solutions obtained by
solving optimization problems. To enhance the accuracy and efficient particle swarm optimizers and genetic algorithms
convergence rate of this algorithm, an improved cuckoo search (Yang 2010). This paper develops an Improved Cuckoo
algorithm is proposed in this paper. Normally, the parameters
Search (ICS) algorithm for unconstrained optimization
of the cuckoo search are kept constant. This may lead to
decreasing the efficiency of the algorithm. To cope with this
problems. To enhance accuracy and convergence rate of the
issue, a proper strategy for tuning the cuckoo search CS, the ICS employs an improved method for generating
parameters is presented. Considering several well-known new solution vectors.
benchmark problems, numerical studies reveal that the
proposed algorithm can find better solutions in comparison
with the solutions obtained by the cuckoo search. Therefore, it 2. Cuckoo search algorithm
is anticipated that the improved cuckoo search algorithm can
successfully be applied to a wide range of optimization To describe the CS more clearly, the breed behavior of
problems. certain cuckoo species is briefly reviewed.
Keywords: Cuckoo search algorithm, global optimization, Lévy
2.1. Cuckoo breeding behaviour
flight, meta-heuristic, tuning

The CS was inspired by the obligate brood parasitism of

some cuckoo species by laying their eggs in the nests of host
1. Introduction birds. Some cuckoos have evolved in such a way that female
parasitic cuckoos can imitate the colors and patterns of the
Because of computational drawbacks of conventional
eggs of a few chosen host species. This reduces the
numerical methods in solving complex optimization
probability of the eggs being abandoned and, therefore,
problems, researchers may have to rely on meta-heuristic
increases their re-productivity (Payne 2005). It is worth
algorithms. Over the last decades, many meta-heuristic
mentioning that several host birds engage direct conflict
algorithms have been successfully applied to various
with intruding cuckoos. In this case, if host birds discover
engineering optimization problems (Sim 2003; Qing 2006;
the eggs are not their own, they will either throw them away
Zhang 2006; Sickel 2007; Sanchis 2008; Marinakis 2008;
or simply abandon their nests and build new ones,
Serrurier 2008). For most complicated real-world
elsewhere.
optimization problems, they have provided better solutions
in comparison with conventional numerical methods. Parasitic cuckoos often choose a nest where the host bird
To imitate natural phenomena, most meta-heuristic just laid its own eggs. In general, the cuckoo eggs hatch
algorithms combine rules and randomness. These slightly earlier than their host eggs. Once the first cuckoo
phenomena include the biological evolutionary processes, chick is hatched, his first instinct action is to evict the host
such as genetic algorithm (GA) (Holland 1975; Goldberg eggs by blindly propelling the eggs out of the nest. This
1989), evolutionary algorithm (Fogel 1996; De Jong 1975) action results in increasing the cuckoo chick’s share of food
and differential evolution (DE) (Storn 1996), animal provided by its host bird (Payne 2005). Moreover, studies
behavior, such as particle swarm optimization (PSO) show that a cuckoo chick can imitate the call of host chicks
(Kennedy 1995), tabu search (Glover 1977) and ant colony to gain access to more feeding opportunity.
algorithm (ACA) (Dorigo 1996), as well as physical
annealing processes, such as simulated annealing (SA) The CS models such breeding behavior and, thus, can be
(Kirkpatrick 1983). applied to various optimization problems. Yang and Deb

http://journals.usb.ac.ir/IJCIT/en-us/MainPage 31
International Journal of Communications and Information Technology, IJCIT IJCIT-2011-Vol.1-No.1 Dec. 2011

(Yang 2009; Yang 2010), discovered that the performance begin

of the CS can be improved by using Lévy Flights instead of Objective function f x , x  ( x ,....., x d )T
simple random walk. Generate initial population of
n host nests x i (i  ,,..., n )
2.2. Lévy Flights While ( t  MaxGeneration ) or (stop criterion)
Get a cuckoo randomly by Lévy flights
In nature, animals search for food in a random or quasi- evaluate its quality / fitness Fi
random manner. Generally, the foraging path of an animal is Choose a nest among n (say. j) randomly
effectively a random walk because the next move is based if (Fi  Fj ),
on both the current location/state and the transition replace j by the new solution;
probability to the next location. The chosen direction end if
implicitly depends on a probability, which can be modeled A fraction ( p a ) of worse nests
mathematically. Various studies have shown that the flight are abandoned and new ones are built;
behavior of many animals and insects demonstrates the Keep the best solutions
typical characteristics of Lévy flights (Brown 2007). A Lévy (or nests with quality solutions);
flight is a random walk in which the step-lengths are Rank the solutions and find the current best
distributed according to a heavy-tailed probability end while
distribution. After a large number of steps, the distance from Postprocess results and visualization
the origin of the random walk tends to a stable distribution. end

2.3. Cuckoo Search Implementation

When generating new solutions x i ( t  ) for the i th
Each egg in a nest represents a solution, and a cuckoo egg
represents a new solution. The aim is to employ the new and cuckoo, the following Lévy flight is performed
potentially better solutions (cuckoos) to replace not-so-good
solutions in the nests. In the simplest form, each nest has x i t    x i t     Lévy   (1)
one egg. The algorithm can be extended to more where    is the step size, which should be related to the
complicated cases in which each nest has multiple eggs
scale of the problem of interest. The product  means
representing a set of solutions (Yang 2009; Yang 2010). The
entry-wise multiplications (Yang 2010). In this research
CS is based on three idealized rules:
work, we consider a Lévy flight in which the step-lengths
• Each cuckoo lays one egg at a time, and dumps it in a are distributed according to the following probability
randomly chosen nest; distribution
Lévy u  t - ,     (2)
• The best nests with high quality of eggs (solutions) will which has an infinite variance. Here, the consecutive
carry over to the next generations; jumps/steps of a cuckoo essentially form a random walk
process which obeys a power-law step-length distribution
• The number of available host nests is fixed, and a host
with a heavy tail.
can discover an alien egg with probability p a  [,] . In It is worth pointing out that, in the real world, if a
this case, the host bird can either throw the egg away or cuckoo’s egg is very similar to a host’s eggs, then this
abandon the nest to build a completely new nest in a new cuckoo’s egg is less likely to be discovered, thus the fitness
location (Yang 2009). should be related to the difference in solutions. Therefore, it
is a good idea to do a random walk in a biased way with
For simplicity, the last assumption can be approximated some random step sizes (Yang 2009).
by a fraction p a of the n nests being replaced by new
nests, having new random solutions. For a maximization 3. Improved Cuckoo Search
problem, the quality or fitness of a solution can simply be
proportional to the objective function. Other forms of fitness
The parameters p a ,  and  introduced in the CS help the
can be defined in a similar way to the fitness function in
genetic algorithms (Yang 2009). algorithm to find globally and locally improved solutions,
respectively. The parameters p a and  are very important
Based on the above-mentioned rules, the basic steps of the parameters in fine-tuning of solution vectors, and can be
CS can be summarized as the pseudo code, as follows (Yang potentially used in adjusting convergence rate of algorithm.
2009):
The traditional CS algorithm uses fixed value for both pa
and  . These values are set in the initialization step and

http://journals.usb.ac.ir/IJCIT/en-us/MainPage 32
International Journal of Communications and Information Technology, IJCIT IJCIT-2011-Vol.1-No.1 Dec. 2011

cannot be changed during new generations. The main parameter “iter”’ refers to the maximum iteration number,
drawback of this method appears in the number of iterations and the parameter “SD” represents the standard deviation.
to find an optimal solution. If the value of p a is small and Considering the best, worst, mean and SD criteria, it can be
seen from Table 3 that all best results for N=10 are given by
the value of  is large, the performance of the algorithm
the ICS.
will be poor and leads to considerable increase in number of
For N=30, Table 4 shows that the ICS leads to better results
iterations. If the value of p a is large and the value of  is than the CS in all criteria, in all test functions except f  , f 
small, the speed of convergence is high but it may be unable
, f  , f , f , and f . For f  , f , and f the CS gives
to find the best solutions.
better results than the ICS in all criteria. For f  and f the
The key difference between the ICS and CS is in the way of worst, mean and SD criteria provided by the ICS is better
adjusting p a and  . To improve the performance of the CS than those given by the CS.
For N=50, Table 5 shows that the ICS leads to better
algorithm and eliminate the drawbacks lies with fixed values
results than the CS in all criteria, in f , f  , f  , f  , f  , f ,
of p a and  , the ICS algorithm uses variables p a and 
f , and f . For f  , f  , f , and f the CS gives
.In the early generations, the values of pa and  must be
better results than the ICS in all criteria. Given the second
big enough to enforce the algorithm to increase the diversity
test function, the worst, mean and SD criteria provided by
of solution vectors. However, these values should be
decreased in final generations to result in a better fine-tuning the ICS is better than those given by the CS. For f  , and
of solution vectors. The values of p a and  are f the ICS manages to better results in terms of the best,
dynamically changed with the number of generation and worst and mean criteria.
expressed in Equations 3-5, where NI and gn are the
number of total iterations and the current iteration, 4.2. Effects of changing the optimization parameters
respectively. on the performance of the ICS
In this subsection, the effect of changing p a and  on the
gn
Pa (gn)  Pa max   Pa max - Pa min  (3)
performance of the ICS is investigated. For N=10, N=30 and
N=50, the iteration number is set to 1000, 3000 and 5000
NI
(gn)   max expc.gn  (4) and the effects of changing p a and  on the performance
of the ICS are shown in Tables 6-8 and Tables 9-11,
  
c Ln min  (5) respectively. Thirty independent runs are carried out in each
NI   max  case and the mean and SD values are obtained.
As can be seen from Tables 6-8, if the minimum value of
p a is decreased with no change in the value of  , better
4. Case studies: analysis and discussion
results may be obtained in most cases. It seems that for test
To verify the reliability of ICS algorithm, several well- functions with high decision variables, an increase in the
known test functions (Zou 2010), as shown in Table 1, are maximum value of p a leads to better results.
considered. In the experiments, the parameters of CS and Tables 9-11 show that a reduction in the minimum value of
ICS algorithms are shown in Table 2, where N is the number  with no change in the value of p a does not have any
of decision variables.
significant effect on the performance of the algorithm.
However, the performance of the algorithm may deteriorate
4.1. Comparison of the ICS and CS algorithms
by an increase in the maximum value of  . Various tests
To optimize the given test functions, a MATLAB code,
show that the suitable algorithm parameters leading to good
using MATLAB Ver.7.10, is developed based on the ICS. In
this version of MATLAB, the numbers smaller than results can approximately be p a min  0.005, p a max  1,
4.9407e-324 are considered as zero. The PC used is an  min =0.05, and  max  . .
INTEL32, X2, 3.0GHz having 4GB of memory. To show
the effectiveness of the proposed algorithm, it is compared
with the CS algorithms. 5. Conclusions
Considering functions f  f , Tables 3-5 show the
optimization results of the implementation of the CS and In this paper, an improved cuckoo search algorithm
ICS algorithms for N=10, 30 and 50. Thirty independent enhancing the accuracy and convergence rate of the cuckoo
experiments are carried out in each case, and the search algorithm was proposed. The impact of keeping the
optimization results are reported. In these tables, the parameters of the cuckoo search algorithm constant was
discussed and a strategy for improving the performance of

http://journals.usb.ac.ir/IJCIT/en-us/MainPage 33
International Journal of Communications and Information Technology, IJCIT IJCIT-2011-Vol.1-No.1 Dec. 2011

the algorithm by properly tuning these parameters was Payne R B, Sorenson M D, Klitz K (2005) The cuckoos. Oxford University
Press.
presented. According to the simulation results, the proposed
Qing AY (2006) Dynamic differential evolution strategy and applications
approach performed well in several benchmark problems in in electromagnetic inverse scattering problems. IEEE T Geosci Remote 44
terms of the accuracy of the solutions found. (1):116–125.
Sanchis J, Martanez MA, Blasco X (2008) Integrated multiobjective
optimization and a priori preferences using genetic algorithms. Inform
References Sciences 178(4):931–951.
Serrurier M, Prade H (2008) Improving inductive logic programming by
Brown C, Liebovitch LS, Glendon R (2007) Lévy flights in Dobe Ju/hoansi using simulated annealing. Inform Sciences 178 (6):1423–1441.
foraging patterns, Hum Ecol. 35:129-138. Sickel JHV, Lee KY, Heo JS (2007) Differential evolution and its
De Jong K (1975) Analysis of the behavior of a class of genetic adaptive applications to power plant control. In: International Conference on
systems, Ph.D. Dissertation, University of Michigan, Ann Arbor, MI. Intelligent Systems Applications to Power Systems, ISAP, 1–6.
Dorigo M, Maniezzo V, Golomi A (1996) Ant system: optimization by a Sim KM, Sun WH (2003) Ant colony optimization for routing and load-
colony of cooperating agents. IEEE T Syst Man Cy B 26 (1) 29–41. balancing: survey and new directions. IEEE T Syst Man Cy A 33 (5):560–
Fogel LJ, Owens AJ, Walsh MJ (1966) Artificial intelligence through 572.
simulated evolution, John Wiley, Chichester, UK. Storn R (1996) Differential evolution design of an IIR-filter. Paper
Glover F (1977) Heuristic for integer programming using surrogate presented in: IEEE International Conference on Evolutionary Computation,
constraints, Decision Sci. 8 (1):156–166. Nagoya, 268–273 1996.
Goldberg DE (1989) Genetic algorithms in search, optimization and Yang XS, Deb S (2009) Cuckoo search via Lévy flights, In: Proceeings of
machine learning, Addison Wesley, Boston, USA. World Congress on Nature & Biologically Inspired Computing (NaBIC
Holland JH (1975) Adaptation in natural and artificial systems. University 2009, India), 210-214.
of Michigan Press, Ann Arbor, USA. Yang XS, Deb S (2010) Engineering Optimisation by Cuckoo Search, Int.
Kennedy J, Eberhart RC (1995) Particle swarm optimization, In: J. Mathematical Modelling and Numerical Optimisation, 1(4):330–343.
Proceedings of IEEE International Conference on Neural Networks. 1942– Zhang JJ, Zhang YH, Gao RZ (2006) Genetic algorithms for optimal design
1948. of vehicle suspensions. Paper presented in IEEE International Conference
Kirkpatrick S, Gelatt C, Vecchi M (1983) Optimization by simulated on Engineering of Intelligent Systems, 1–6.
annealing, Sciene. 220:671–680. Zou DX, et al., (2010) Novel global harmony search algorithm for
Koza JR (1990) Genetic programming: a paradigm for genetically breeding unconstrained problems, Neurocomputing. 73:3308-3318.
populations of computer programs to solve problems,Stanford University,
USA.
Marinakis Y, Marinaki M, Dounias G (2008) Particle swarm optimization
for pap- smear diagnosis. Expert Syst Appl 35 (4):1645–1656.

http://journals.usb.ac.ir/IJCIT/en-us/MainPage 34
International Journal of Communications and Information Technology, IJCIT IJCIT-2011-Vol.1-No.1 Dec. 2011

Table1: Benchmark problems

Name Test Function Search Space Optimum

f (Sphere function) N [-100,100]n 0

min f   x i
i 
f  (Rosenbrock
function)
N 

 
min f     x i   x i  x i   
i   
 [-100,100]n 0

f  (Generalized
Rastrigrin function)
N

min f    x i   cosx i   
i 
 [-10,10]n 0

f  (Generalized  N  N  xi  [-600,600]n 0
min f    x i   cos   
Griewank function)  i  i   i
f  (Ackley’s    [-32,32]n 0
 iN x i 
function) min f     e   exp  . 
 N 
 
  N cosx i  
 exp i  
 N 
 
f  (Schwefel’s N N [-100,100]n 0
min f    x i   x i
problem 2.22) i  i 
f  (Schwefel’s min f   . N   x i sin
N

i 
 x i
[-500,500]n n=10,
1.2728e-04
problem 2.26)
n=30,
3.8183e-04
n=50,
6.3638e-04
f  (Rotated hyper- 2 [-100,100]n 0
 i  N

ellipsoid function) min f8     x j 

i 1  j 1 
f N [-100,100]n -450
min f    z i  
i 
f  [-100,100]n -450
N i
min f     z j   
 
i  j 

f  [-100,100]n -450

N i 
min f     z j    . * N,   
 
i  j 

f N 

i  

 
min f    zi   zi  zi     

 [-100,100]n 390

f  N  N z  [-600,600]n -180

min f   z i   cos i     
 i  i   i
f    [-32,32]n -140
 iN zi 
min f    e   exp  . 
 N 
 
  N coszi  
 exp i    
 N 
 
f N

min f   z i   cos z i     
i 
 [-5,5]n -330

http://journals.usb.ac.ir/IJCIT/en-us/MainPage 35
International Journal of Communications and Information Technology, IJCIT IJCIT-2011-Vol.1-No.1 Dec. 2011

Table 2: Algorithms’ parameters

Algorithms Number of Number of pa  
Dimension Generation
CS N=10 1000 0.1 1.5 0.25
N=30 3000
N=50 5000
ICS N=10 1000 p a (max)  . 1.5 (max)  .
N=30 3000 p a (min)  . (min)  .
N=50 5000

Table 3: The optimization results of the CS and ICS algorithms for f  f (N=10)
Function Algorithm iter Best Worst Mean SD
f CS 1000 4.9826e-18 1.5338e-15 2.5086e-16 3.5572e-16
ICS 1000 2.3482e-25 5.9057e-21 4.2599e-22 1.1657e-21

f CS 1000 7.5742e-01 7.4146e+01 5.1543e+00 1.3235e+01

ICS 1000 2.3671e-02 1.1031e+01 3.7763e+00 3.2777e+00

f CS 1000 2.4935e+00 1.1197e+01 5.8173e+00 2.2131e+00

ICS 1000 1.4728e+00 5.0832e+00 3.1597e+00 9.1871e-01

f CS 1000 1.2824e-02 1.0010e-01 5.3369e-02 2.1703e-02

ICS 1000 2.7491e-03 6.3613e-02 3.4795e-02 1.6791e-02

f CS 1000 3.1724e-08 1.1169e-05 1.1978e-06 2.4007e-06

ICS 1000 2.1651e-11 3.4678e-06 1.5947e-07 6.5468e-07

f CS 1000 8.7718e-06 3.9122e-03 2.5590e-04 7.0628e-04

ICS 1000 1.8418e-11 1.2817e-08 8.0449e-10 2.3225e-09

f CS 1000 2.0399e+02 8.1834e+02 4.7802e+02 1.8304e+02

ICS 1000 7.1429e-01 6.9924e+02 3.8381e+02 1.6208e+02

f CS 1000 8.5143e-09 1.6294e-06 2.3584e-07 3.5945e-07

ICS 1000 3.5209e-11 1.0386e-07 1.5240e-08 2.6462e-08

f CS 1000 -450 -450 -450 2.9856e-14

ICS 1000 -450 -450 -450 1.8283e-14

f CS 1000 -450 -4.5000e+02 -450 3.7084e-07

ICS 1000 -450 -450 -450 6.2205e-08

f CS 1000 -4.5000e+02 -4.4990e+02 -4.4999e+02 2.0502e-02

ICS 1000 -450 -4.5000e+02 -4.5000e+02 4.2198e-04

f CS 1000 3.9031e+02 4.7311e+02 3.9788e+02 1.5458e+02

ICS 1000 3.9011e+02 4.0259e+02 3.9329e+02 2.9123e+00

f CS 1000 -1.7998e+02 -1.7992e+02 -1.7995e+02 1.5447e-02

ICS 1000 -1.8000e+02 -1.7993e+02 -1.7996e+02 1.4554e-02

f CS 1000 -140 -1.3884e+02 -1.3996e+02 2.1090e-01

ICS 1000 -140 -140 -140 2.9775e-07

f CS 1000 -3.2848e+02 -3.1921e+02 -3.2422e+02 2.2932e+00

ICS 1000 -3.2979e+02 -3.2421e+02 -3.2698e+02 1.0785e+00

http://journals.usb.ac.ir/IJCIT/en-us/MainPage 36
International Journal of Communications and Information Technology, IJCIT IJCIT-2011-Vol.1-No.1 Dec. 2011

Table 4: The optimization results of the CS and ICS algorithms for f  f (N=30)
Function Algorithm iter Best Worst Mean SD
f CS 3000 1.2421e-15 7.3248e-13 6.6222e-14 1.3432e-13
ICS 3000 2.9807e-22 4.1015e-20 9.5438e-21 1.1279e-20

f CS 3000 1.8759e+00 1.5392e+02 4.0384e+01 3.5250e+01

ICS 3000 9.8653e+00 7.9836e+01 2.6678e+01 1.3697e+01

f CS 3000 2.0423e+01 5.9923e+01 3.6513e+01 9.2702e+00

ICS 3000 1.1939e+01 2.9747e+01 2.2296e+01 4.1242e+00

f CS 3000 1.6098e-14 7.5394e-02 1.0846e-02 2.0119e-02

ICS 3000 1.1102e-16 4.9058e-08 3.1173e-09 1.1340e-08

f CS 3000 5.7283e-07 4.3404e+00 2.0926e+00 8.4421e-01

ICS 3000 7.0179e-10 2.0393e+00 3.0880e-01 5.9632e-01

f CS 3000 1.4660e-05 9.8736e+01 3.7578e+00 1.7985e+01

ICS 3000 2.1208e-10 2.8235e-06 2.1058e-07 5.4655e-07

f CS 3000 1.4402e+03 3.1537e+03 2.4094e+03 4.3460e+02

ICS 3000 1.4154e+03 3.3428e+03 2.5541e+03 4.1141e+02

f CS 3000 7.9269e-01 8.0358e+00 2.8559e+00 1.6281e+00

ICS 3000 1.1422e+00 1.5802e+01 5.2848e+00 3.4241e+00

f CS 3000 -450 -450 -450 2.3366e-13

ICS 3000 -450 -450 -450 1.0449e-13

f CS 3000 -4.4936e+02 -4.4103e+02 -4.4762e+02 1.8178e+00

ICS 3000 -4.4847e+02 -4.0914e+02 -4.4034e+02 8.6941e+00

f CS 3000 -2.3566e+02 6.9350e+02 2.4178e+02 2.8887e+02

ICS 3000 -1.5538e+02 1.3238e+03 3.7542e+02 3.3414e+02

f CS 3000 3.9042e+02 5.6100e+02 4.4837e+02 4.2947e+01

ICS 3000 3.9679e+02 4.7957e+02 4.1953e+02 2.1417e+01

f CS 3000 -1.8000e+02 -1.7997e+02 -1.8000e+02 7.9881e-03

ICS 3000 -180 -180 -180 2.0596e-10

f CS 3000 -1.4000e+02 -1.3373e+02 -1.3786e+02 1.0946e+00

ICS 3000 -140 -1.3577e+02 -1.3978e+02 8.2844e-01

f CS 3000 -3.1378e+02 -2.8555e+02 -3.0235e+02 7.0745e+00

ICS 3000 -3.1666e+02 -2.9349e+02 -3.0686e+02 5.5904e+00

http://journals.usb.ac.ir/IJCIT/en-us/MainPage 37
International Journal of Communications and Information Technology, IJCIT IJCIT-2011-Vol.1-No.1 Dec. 2011

Table 5: The optimization results of the CS and ICS algorithms for f  f (N=50)
Function Algorithm iter Best Worst Mean SD
f CS 5000 5.4547e-13 4.9841e-10 2.7691e-11 9.0314e-11
ICS 5000 1.6080e-20 1.1525e-18 2.0675e-19 2.3478e-19

f CS 5000 9.8038e+00 7.3955e+02 1.5139e+02 1.2737e+02

ICS 5000 3.3689e+01 1.7960e+02 7.6080e+01 3.9611e+01

f CS 5000 4.6418e+01 1.0788e+02 7.1486e+01 1.3445e+01

ICS 5000 3.3463e+01 7.1683e+01 5.3111e+01 1.0093e+01

f CS 5000 1.9523e-12 5.1408e-02 8.5133e-03 1.4217e-02

ICS 5000 3.3307e-16 3.6921e-09 1.2358e-10 6.7400e-10

f CS 5000 2.2001e+00 6.8027e+00 4.0143e+00 1.0781e+00

ICS 5000 8.2101e-09 3.5109e+00 8.0832e-01 1.2424e+00

f CS 5000 8.8532e-04 2.4962e+02 2.7147e+01 6.2970e+01

ICS 5000 3.9764e-10 8.7367e-05 3.8818e-06 1.59140e-05

f CS 5000 2.7726e+03 6.0688e+03 4.4152e+03 7.2422e+02

ICS 5000 2.9766e+03 6.1253e+03 4.7645e+03 7.6395e+02

f CS 5000 2.9885e+01 1.9126e+02 9.1219e+01 4.4384e+01

ICS 5000 3.4801e+01 3.5944e+02 1.7395e+02 7.4926e+01

f CS 5000 -450 -450 -450 4.1492e-11

ICS 5000 -450 -450 -450 2.2466e-13

f CS 5000 -4.2126e+02 -3.1512e+02 -3.7815e+02 3.0840e+01

ICS 5000 -3.9594e+02 -6.3480e+01 -2.4841e+02 7.1120e+01

f CS 5000 1.7480e+03 5.9222e+03 3.1950e+03 9.4380e+02

ICS 5000 2.0394e+03 6.1629e+03 4.1276e+03 1.1680e+03

f CS 5000 4.1866e+02 1.2019e+03 5.6663e+02 1.5325e+02

ICS 5000 4.1500e+02 5.5800e+02 4.6508e+02 4.2375e+01

f CS 5000 -180 -1.7991e+02 -1.7999e+02 2.1715e-02

ICS 5000 -180 -180 -180 2.7292e-12

f CS 5000 -1.3805e+02 -1.3008e+02 -1.3577e+02 1.3777e+00

ICS 5000 -1.4000e+02 -1.3107e+02 -1.3897e+02 2.0029e+00

f CS 5000 -2.9677e+02 -2.4743e+02 -2.7692e+02 1.1287e+01

ICS 5000 -2.9753e+02 -2.6720e+02 -2.8149e+02 6.5472e+01

http://journals.usb.ac.ir/IJCIT/en-us/MainPage 38
International Journal of Communications and Information Technology, IJCIT IJCIT-2011-Vol.1-No.1 Dec. 2011

Table 6: The effect of changing p a with no change in  on the performance of the ICS (N=10)
Function Pa  .  . Pa    . Pa  .  .
f Mean 4.2599e-22 4.4653e-25 1.0340e-26
SD 1.1657e-21 8.3606e-25 2.5207e-26
f Mean 3.7763e+00 4.9089e+00 2.3611e+00
SD 3.2777e+00 5.5049e+00 3.1290e+00
f Mean 3.1597e+00 1.8303e+00 2.8445e+00
SD 9.1871e-01 9.5162e-01 1.7087e+00
f Mean 3.4795e-02 3.8763e-02 3.9614e-02
SD 1.6791e-02 1.8304e-02 2.1058e-02
f Mean 1.5947e-07 1.0920e-09 3.6240e-10
SD 6.5468e-07 2.8637e-09 1.7380e-09
f Mean 8.0449e-10 8.2101e-13 1.3079e-12
SD 2.3225e-09 1.2657e-12 2.9891e-12
f Mean 3.8381e+02 2.5381e+02 3.3117e+02
SD 1.6208e+02 1.0782e+02 1.4635e+02
f Mean 1.5240e-08 1.5016e-06 1.3773e-09
SD 2.6462e-08 2.9948e-06 2.1917e-09
f Mean -450 -450 -450
SD 1.8283e-14 0 1.4928e-14
f Mean -450 -450 -450
SD 6.2205e-08 2.3677e-06 3.3367e-08
f Mean -4.5000e+02 -4.4999e+02 -4.5000e+02
SD 4.2198e-04 1.1282e-02 1.1264e-04
f Mean 3.9329e+02 3.9846e+02 3.9247e+02
SD 2.9123e+00 1.1820e+02 2.4899e+00
f Mean -1.8000e+02 -1.7996e+02 -1.7996e+02
SD 1.4554e-02 1.9431e-02 2.0279e-02
f Mean -140 -1.4000e+02 -140
SD 2.9775e-07 1.3034e-03 5.9855e-10
f Mean -3.2698e+02 -3.2831e+02 -3.2726e+02
SD 1.0785e+00 7.4429e-01 1.3477e+00

http://journals.usb.ac.ir/IJCIT/en-us/MainPage 39
International Journal of Communications and Information Technology, IJCIT IJCIT-2011-Vol.1-No.1 Dec. 2011

Table 7: The effect of changing p a with no change in  on the performance of the ICS (N=30)
Function Pa  .  . Pa    . Pa  .  .
f Mean 9.5438e-21 4.6272e-25 2.0677e-24
SD 1.1279e-20 5.1117e-25 5.6379e-24
f Mean 2.6678e+01 3.3602e+01 3.1223e+01
SD 1.3697e+01 1.9880e+01 2.2979e+01
f Mean 2.2296e+01 2.9557e+01 2.5750e+01
SD 4.1242e+00 4.1747e+00 6.9751e+00
f Mean 3.1173e-09 4.7222e-09 2.1770e-07
SD 1.1340e-08 2.1088e-08 8.2739e-07
f Mean 3.0880e-01 1.4052e-01 3.8519e-02
SD 5.9632e-01 5.3836e-01 2.1096e-01
f Mean 2.1058e-07 1.0914e-12 1.2310e-10
SD 5.4655e-07 3.1045e-12 4.4570e-10
f Mean 2.5541e+03 2.2606e+03 2.4286e+03
SD 4.1141e+02 3.6773e+02 5.7437e+02
f Mean 5.2848e+00 4.8976e+01 4.7942e+00
SD 3.4241e+00 2.5553e+01 3.4457e+00
f Mean -450 -450 -450
SD 1.0449e-13 2.1111e-14 6.9217e-14
f Mean -4.4034e+02 -3.9115e+02 -4.4513e+02
SD 8.6941e+00 4.4652e+02 3.3119e+00
f Mean 3.7542e+02 1.0171e+03 2.8481e+02
SD 3.3414e+02 5.4128e+02 3.7925e+02
f Mean 4.1953e+02 4.2623e+02 4.2345e+02
SD 2.1417e+01 2.5360e+01 2.4905e+01
f Mean -180 -180 -180
SD 2.0596e-10 9.7676e-07 1.6358e-07
f Mean -1.3978e+02 -1.400e+02 -1.3962e+02
SD 8.2844e-01 8.2856e-03 2.1090e-01
f Mean -3.0686e+02 -3.0164e+02 -3.0661e+02
SD 5.5904e+00 5.4853e+00 6.8482e+00

http://journals.usb.ac.ir/IJCIT/en-us/MainPage 40
International Journal of Communications and Information Technology, IJCIT IJCIT-2011-Vol.1-No.1 Dec. 2011

Table 8: The effect of changing p a with no change in  on the performance of the ICS (N=50)
Function Pa  .  . Pa    . Pa  .  .
f Mean 2.0675e-19 3.1272e-23 9.0087e-24
SD 2.3478e-19 4.3418e-23 1.0870e-23
f Mean 7.6080e+01 6.7831e+01 7.9552e+01
SD 3.9611e+01 2.9552e+01 5.2299e+01
f Mean 5.3111e+01 6.2762e+01 5.8441e+01
SD 1.0093e+01 1.3800e+01 1.1241e+01
f Mean 1.2358e-10 8.9732e-14 2.4653e-04
SD 6.7400e-10 4.0347e-13 1.3503e-03
f Mean 8.0832e-01 1.8687e+00 3.5381e-01
SD 1.2424e+00 4.9738e+00 9.2685e-01
f Mean 3.8818e-06 7.4982e-12 3.0294e-10
SD 1.59140e-05 1.1453e-11 1.0090e-09
f Mean 4.7645e+03 4.6308e+03 4.4703e+03
SD 7.6395e+02 6.1886e+02 9.6839e+02
f Mean 1.7395e+02 8.8711e+02 1.1761e+01
SD 7.4926e+01 2.5878e+02 6.0230e+01
f Mean -450 -450 -450
SD 2.2466e-13 1.2035e-13 1.2534e-13
f Mean -2.4841e+02 4.9511e+02 -3.1568e+02
SD 7.1120e+01 2.6049e+02 7.8122e+01
f Mean 4.1276e+03 7.1806e+03 3.5103e+02
SD 1.1680e+03 2.4540e+03 1.3683e+02
f Mean 4.6508e+02 4.4468e+02 4.6210e+02
SD 4.2375e+01 2.0388e+01 3.8253e+01
f Mean -180 -180 -1.800e+02
SD 2.7292e-12 2.6177e-13 2.5572e-03
f Mean -1.3897e+02 -1.3956e+02 -1.3963e+02
SD 2.0029e+00 8.6614e+01 8.6484-01
f Mean -2.8149e+02 -2.6482e+02 -2.8186e+02
SD 6.5472e+01 1.2553e+01 9.0449e+00

http://journals.usb.ac.ir/IJCIT/en-us/MainPage 41
International Journal of Communications and Information Technology, IJCIT IJCIT-2011-Vol.1-No.1 Dec. 2011

Table 9: The effect of changing  with no change in p a on the performance of the ICS (N=10)
Function   0. - 0.01   1 - 0.001   0.5 - 0.001
f Mean 4.2599e-22 7.0038e-23 8.9021e-20
SD 1.1657e-21 1.6336e-22 3.5120e-19
f Mean 3.7763e+00 5.9300e+00 8.6980e+00
SD 3.2777e+00 1.2964e+01 1.7134e+00
f Mean 3.1597e+00 3.3373e+00 3.9880e+00
SD 9.1871e-01 1.3127e+00 1.3427e+00
f Mean 3.4795e-02 4.0566e-02 3.2937e-02
SD 1.6791e-02 1.8490e-02 1.3853e-02
f Mean 1.5947e-07 3.9132e-06 3.7450e-06
SD 6.5468e-07 1.8086e-05 1.9565e-05
f Mean 8.0449e-10 7.7525e-11 2.7757e-08
SD 2.3225e-09 1.0322e-10 8.0469e-08
f Mean 3.8381e+02 3.3703e+02 4.0397e+02
SD 1.6208e+02 1.6933e+02 1.4764e+02
f Mean 1.5240e-08 8.0830e-08 1.1071e-07
SD 2.6462e-08 1.5611e-07 1.4769e-07
f Mean -450 -450 -450
SD 1.8283e-14 1.0556e-14 1.8283e-14
f Mean -450 -450 -450
SD 6.2205e-08 4.9119e-07 1.7472e-07
f Mean -4.5000e+02 -4.5000e+02 -4.500e+02
SD 4.2198e-04 3.5347e-02 1.4650e-03
f Mean 3.9329e+02 3.9606e+02 4.0136e+02
SD 2.9123e+00 7.4060e+00 2.0514e+01
f Mean -1.8000e+02 -1.7996e+02 -1.7996e+02
SD 1.4554e-02 1.6862e-02 1.7038e-02
f Mean -140 -1.4000e+02 -140
SD 2.9775e-07 4.6891e-05 1.1664e-06
f Mean -3.2698e+02 -3.2648e+02 -3.2606e+02
SD 1.0785e+00 1.1100e+00 1.7840e+00

http://journals.usb.ac.ir/IJCIT/en-us/MainPage 42
International Journal of Communications and Information Technology, IJCIT IJCIT-2011-Vol.1-No.1 Dec. 2011

Table 10: The effect of changing  with no change in p a on the performance of the ICS (N=30)
Function   0. - 0.01   1 - 0.001   0.5 - 0.001
f Mean 9.5438e-21 4.6000e-20 1.2784e-17
SD 1.1279e-20 1.3378e-19 2.1359e-17
f Mean 2.6678e+01 2.5784e+01 3.0142e+01
SD 1.3697e+01 1.5047e+01 2.5385e+01
f Mean 2.2296e+01 2.4978e+01 2.4610e+01
SD 4.1242e+00 5.4647e+00 4.6284e+00
f Mean 3.1173e-09 2.4654e-04 1.2254e-06
SD 1.1340e-08 1.3503e-03 4.7742e-06
f Mean 3.0880e-01 2.3712e+00 5.8888e-01
SD 5.9632e-01 3.8305e+00 1.1604e+00
f Mean 2.1058e-07 6.8962e-08 6.9028e-06
SD 5.4655e-07 1.2212e-07 1.9754e-05
f Mean 2.5541e+03 2.6420e+03 2.6912e+03
SD 4.1141e+02 4.3081e+02 4.2520e+02
f Mean 5.2848e+00 6.1728e+00 1.2693e+01
SD 3.4241e+00 4.6207e+00 1.2218e+01
f Mean -450 -450 -450
SD 1.0449e-13 1.1221e-13 1.6656e-13
f Mean -4.4034e+02 -4.4102e+02 -4.4207e+02
SD 8.6941e+00 6.8546e+00 7.1356e+00
f Mean 3.7542e+02 3.6437e+02 4.0537e+02
SD 3.3414e+02 3.4552e+02 4.5036e+02
f Mean 4.1953e+02 4.2239e+02 4.2657e+02
SD 2.1417e+01 2.5361e+01 2.4936e+01
f Mean -180 -1.8000e+02 -1.8000e+02
SD 2.0596e-10 1.8763e-03 1.3503e-03
f Mean -1.3978e+02 -1.3649e+02 -1.3985e+02
SD 8.2844e-01 5.3672e+00 6.8264e-01
f Mean -3.0686e+02 -3.0725e+02 -3.0540e+02
SD 5.5904e+00 6.2390e+00 6.0573e+00

http://journals.usb.ac.ir/IJCIT/en-us/MainPage 43
International Journal of Communications and Information Technology, IJCIT IJCIT-2011-Vol.1-No.1 Dec. 2011

Table 11: The effect of changing  with no change in p a on the performance of the ICS (N=50)
Function   0. - 0.01   1 - 0.001   0.5 - 0.001
f Mean 2.0675e-19 6.8483e-19 2.2670e-16
SD 2.3478e-19 1.2882e-18 4.8857e-16
f Mean 7.6080e+01 7.6643e+01 7.5390e+01
SD 3.9611e+01 3.0751e+01 3.6539e+01
f Mean 5.3111e+01 5.9818e+01 5.3635e+01
SD 1.0093e+01 1.1159e+01 9.6758e+00
f Mean 1.2358e-10 9.8614e-04 3.8313e-10
SD 6.7400e-10 2.5572e-03 1.6581e-09
f Mean 8.0832e-01 7.7256e+00 1.4874e+00
SD 1.2424e+00 7.2623e+00 1.8926e+00
f Mean 3.8818e-06 1.1599e-07 1.0947e-04
SD 1.5914e-05 2.1111e-07 1.9751e-04
f Mean 4.7645e+03 5.0340e+03 4.9477e+03
SD 7.6395e+02 9.6688e+02 7.9346e+02
f Mean 1.7395e+02 2.2547e+02 2.1874e+02
SD 7.4926e+01 9.8939e+01 8.1697e+01
f Mean -450 -450 -450
SD 2.2466e-13 2.1399e-13 2.9366e-13
f Mean -2.4841e+02 -2.4571e+02 -2.0653e+02
SD 7.1120e+01 7.5262e+01 8.7765e+01
f Mean 4.1276e+03 4.2418e+03 4.2070e+03
SD 1.1680e+03 1.5940e+03 1.5753e+03
f Mean 4.6508e+02 4.6803e+02 4.6397e+02
SD 4.2375e+01 3.7100e+01 3.2449e+01
f Mean -180 -1.8000e+02 -1.8000e+02
SD 2.7292e-12 3.7436e-03 3.3796e-03
f Mean -1.3897e+02 -1.3338e+02 -1.3727e+02
SD 2.0029e+00 7.1231e+00 3.88013+00
f Mean -2.8149e+02 -2.8409e+02 -2.8475e+02
SD 6.5472e+01 1.0044e+01 9.2662e+00

http://journals.usb.ac.ir/IJCIT/en-us/MainPage 44