Documentos de Académico
Documentos de Profesional
Documentos de Cultura
Abstract Digital infinite impulse response (IIR) filter have based algorithms, the ICA starts with a random generated
become the target of growing interest, because they often provide population called countries. The countries with best costs are
a much better performance and less computational cost than considered as imperialists and the rest of them are called
finite impulse response (FIR) filters. Since the problem of error colonies. Movement of the colonies toward their related
surface of designing Digital IIR filters is generally nonlinear and
imperialist and imperialistic competition among the empires
multimodal, global optimization techniques are required in order
to avoid local minima. In this paper, an evolutionary method form the basis of the ICA. These mechanisms make population
based on Imperialist Competitive Algorithm (ICA) has been to converge to certain spots in the search space where the best
proposed to design Digital IIR filters. By adding a step to the solution founded so far are located.
standard ICA algorithm, its performance has been improved in Exploration and exploitation are two major concepts in
searching solution space and convergence to the global minima. evolutionary algorithms. More exploration results in creating
Simulation results show the efficiency of the proposed method to more solution candidate in the search space which helps the
design Digital IIR filter. algorithm to escape from local minima. On the contrary,
exploitation improves existing solutions and results in fast
Index Terms IIR digital filter, Adaptive filter, System
identification, Optimization, Imperialist Competitive Algorithm. convergence in these algorithms. For instance, in GA,
crossover and mutation operators provide exploitation and
exploration, respectively. Also, lower and upper values for
learning factors result in better exploitation and exploration in
I.INTRODUCTION
PSO algorithm, respectively. In addition, appropriate
Digital IIR filters are useful in a wide range of applications modification in evolutionary algorithms can improve
where a highly-selective processing of discrete signals is exploration and exploitation concepts. Changing operators and
desirable. They are used to modify, reshape, or manipulate the steps and also adding some steps to the existing steps are
frequency spectrum of a signal according to some desired examples of appropriate modification in these algorithms [8-
specifications. These filters often provide a much better 10].
performance and less computational cost than FIR filters [1-3]. In this paper, performance of standard ICA has been
A major concern in IIR filter design is that the cost function improved by an additional step. This modification improves
of design problem has generally multimodal error surface. This exploration and exploitation and results in better balance
multimodal error surface causes the design process to get stuck between these two concepts. Application of the modified ICA
in local minima easily and prevents it to converge to global (MICA) for designing digital IIR filters shows the performance
optimum. In order to deal with this problem, a design method of MICA to find better solution and escape from local minima.
is required to achieve global minima in a multi-modal error The remainder of the paper is organized as follows. The
surface. In recent years, evolutionary algorithms have been problem of digital IIR filter design is described in section 2. In
employed to find global optimum and scape from local section 3, the standard ICA is introduced and then MICA is
minimum in a nonlinear searching space for multimodal presented. Section 4 investigates the application of this
optimization problem and adaptive filtering. Some employed modified algorithm to design digital IIR filter. Then, the
ones include genetic algorithm (GA) [4], harmony search simulation results are compared to the results obtained from
algorithm (HS) [5], particle swarm optimization algorithm other optimization algorithms such as differential evolution
(PSO) [6] and ICA [7]. (DE), PSO, standard ICA, orthogonal ICA (OICA) [11] and
ICA is a meta-heuristic algorithm which was inspired by ICA combined with GA (GAICA) [9].
the process of socio-political evolution. Like other population
,(((
II.STATEMENT OF THE PROBLEM based on the probability of each one. The initial countries are
The relation between input, output and the transfer formed according to Eq. 4.
function of digital IIR filter is as the following:
x = xmin + (xmax xmin ) rand (1, nvar) (4)
y(n) = U
k=0 bk x nk - V
k=0 ak y nk (1)
where xmin and xmax are lower and upper limits,
respectively. nvar is the number of problem parameters that is
U
k=0 bk Z
-k
H(z) = (2) going to be determined. rand (1, nvar) is a random vector.
-k
1+ V
k=1 ak Z Forming initial empires, imperialists started to improve their
colonies. This fact has been modeled by moving all colonies
where ak and bk are filter coefficients, x(n) and y(n) are toward their imperialist. Figure 2 shows the colonies
input and output, respectively. In digital IIR filter design, the movement toward their imperialist. Also, the relation of this
desired system is unknown. Therefore, the design process is movement is shown in Eq. 5.
defined as a system identification problem as shown in Fig. 1
[12 -14]. In fact, digital IIR filter designing problem is
x' = x + {r} (t x)
introduced as minimization of a cost function defined as mean
squared error (MSE) according to Eq. 3. When this error
Where is greater than 1 and close to 2; > 2 values cause
function reaches its minimum value, optimum weights are
instability [7]. {r} is a random vector and its dimensions is
obtained for the filter.
equal to the problem dimension. It is used to cause deviation on
1 2 the way of movement. t and x are the imperialist and colony
J= kN= 1 d(k) y(k) (3) positions respectively.
N
Fig. 2. Movement of colonies to its new location in the standard ICA [7].
Fig. 1. Block diagram of system identification process using adaptive IIR The next step in revolution. It causes a country to have a
filter. suddenly change in its socio-political characteristics. It is
modeled by random movement of a colony and getting placed
at a new random position. This random movement results in
III.MODIFIED IMPERIALIST COMPETITIVE ALGORITHM exploration and creating new solution candidates to escape
from local minima. Figure 3 describes the random changes in
A. Standard imperialist competitive algorithm colonies position. Equation 6 shows how revolution changes
ICA, introduced by Atashpaz and Lucas [7], is a new colonies position.
evolutionary optimization method which is inspired by
imperialistic competition. Movement of the colonies toward x ~ N(x, ) ~ x + N(0 , 1) (6)
their related imperialist and imperialistic competition among
the empires form basis of the ICA. These mechanisms cause where N(0,1) is a random number between 0 and 1 with
initial population to converge to certain spots in the search normal distribution. is the step length and its value is often
space where the best solution founded so far are located. considered to be equal to (xmax xmin) where is almost
Like other evolutionary algorithms, it starts with an initial equal to 0.1 [7].
population called country. Then, this population is divided into After revolution step, the position of a colony is checked in
two types of colonies and imperialists which all together form each empire. If a colony has better position than imperialist, the
empires. Some countries with lower cost are determined as imperialist country is replaced by the colony.
imperialist and other countries are allocated to each imperialist
learning factors a1 and a2 are set 1.5 and 2.5 respectively.
Although this relation is almost like particle velocity relation in
PSO algorithm, there are some differences between them. In
PSO, particle velocity is related to best position of the particle
and the best position of whole particle swarms' best position.
But, in our modified ICA according to the Eq. 8, new position
of each colony is related to the vectors between the colony
toward its imperialist and the best imperialist.
As the assimilation step causes population convergence to
location of the best solution founded so far in the search space,
additional step in MICA improves the colonies obtained from
assimilation and revolution and converges them to better
solution found so far. Also, choosing appropriate values for a1
and a2 provides better exploration and exploitation to escape
from local minima and converge to global minimum. Figure 4
Fig. 3. Revolution in colonies [10]. shows the colony movement toward its new location in
additional step. Pseudo-code of MICA is shown in Fig. 5.
Then, total power of each empire is calculated as the
imperialist power plus a percentage of the colonies total power
as shown in Eq. 7.
TABLE I. RESULTS OF DIFFERENT ALGORITHMS FOR EXAMPLE 1.
TABLE II. RESULTS OF DIFFERENT ALGORITHMS FOR EXAMPLE 2. TABLE III. RESULTS OF DIFFERENT ALGORITHMS FOR EXAMPLE 3.
REFERENCES [9] M.H. Moradi, A. Zeinalzadeh, Y. Mohammadi and M. Abedini,
"An efficient hybrid method for solving the optimal sitting and
[1] A. Kalinli and N. Karaboga, A new method for adaptive IIR
sizing problem of DG and shunt capacitor banks simultaneously
filter design based on Tabu search algorithm, Int. J. Electron.
based on imperialist competitive algorithm and genetic
Commun. (AE), vol. 59, no. 2, pp. 111117, 2005.
algorithm", International Journal of Electrical Power & Energy
[2] N. Karaboga, A new design method based on artificial bee Systems, vol. 54, pp.101-111, January 2014.
colony algorithm for digital IIR filters, Journal of the Franklin
[10]M. Sharifi, H. Mojallali, Design of IIR Digital Filter using
Institute 346, pp. 328-348, November 2009.
modified chaotic orthogonal imperialist competitive algorithm,
[3] Ch. Dai & W. Weirong, Seeker Optimization Algorithm for International Journal of Engineering, Transactions A: Basics,
Digital IIR Filter Design, IEEE Transactions on industrial Vol. 27, No. 7, pp. 1035- 1044, 2014.
electronics, Vol. 57, no. 5, may 2010.
[11]A. Kaveh, S. Talatahari, "Optimum design of skeletal structures
[4] S.C. NG, S.H. LEUNG, C.Y.CHUNG, A. LUK & W.H. LAU, using imperialist competitive algorithm", journal ScienceDirect,
The Genetic Search Approach. A new Learning algorithm for Computers and Structures, vol. 88, pp.12201229, 2010.
Adaptive IIR filtering, Signal Processing Magazine, IEEE, vol.
[12]A. Kalinli and N. Karaboga, Artificial immune algorithm for
13, Issue 6, pp. 38 46, November 1996.
IIR filter design, J. Eng. Appl. Artif. Intell., vol. 18, no. 5, pp.
[5] S. Ghosh, D. Kundu, K. Suresh, S. Das and A. Abraham, 919929, December. 2005.
Design of Optimal Digital IIR Filters by Using a Bandwidth
[13]J. Luukko, K. Rauma, "Open-loop adaptive filter for power
Adaptive Harmony Search Algorithm, World Congress on
electronics applications", IEEE Transactions on Industrial
Nature & Biologically Inspired Computing (NaBIC), pp. 481-
Electronics, vol. 55 (2), pp. 910917, February 2008.
486, December 2009.
[14]D. J. Krusienski and W. K. Jenkins, Particle swarm
[6] R. Singh and S. K. Arya, Optimization of IIR digital filters
optimization for adaptive IIR filter structures, in Proc. Congr.
using Particle Swarm Optimization, International Conference
Evol.Comput, pp. 965970, January 2004.
on Communication, Information & Computing Technology
(ICCICT), pp. 1 7, October 2012. [15]M. S. White & S. J. Flockton, Adaptive recursive filtering
using evolutionary algorithms, in Evolutionary Algorithms in
[7] E. Atashpaz-Gargari, C. Lucas, Imperialist competitive
Engineering Applications, D. Dasgupta and Z. Michalewicz,
algorithm: an algorithm for optimization inspired by
Eds. Berlin, Germany: Springer-Verlag, 1997, pp. 361376.
imperialistic competition, IEE Cong Evol. Comput, pp.46617,
2007.
[8] S. Talatahari, B. Farahmand Azar , R. Sheikholeslami and A.H.
Gandomi, Imperialist competitive algorithm combined with
chaos for global optimization, Commun Nonlinear Sci. Numer.
Simulat, vol. 17, pp. 1312-1319, March 2012.