See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/314523255

Harmony Search Algorithm

Chapter · March 2017

DOI: 10.4018/978-1-5225-2322-2.ch001

CITATION READS

1 3,945

2 authors:

Alireza Askarzadeh Esmat Rashedi

Kerman Graduate University of Technology Kerman Graduate University of Technology
67 PUBLICATIONS   2,913 CITATIONS    51 PUBLICATIONS   4,895 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Proposing a new class of methaheuristic based on the gravitay laws View project

Prediction of scour around hydaulic structure View project

All content following this page was uploaded by Alireza Askarzadeh on 30 April 2017.

The user has requested enhancement of the downloaded file.

 1

Chapter 1
Harmony Search
Algorithm:
Basic Concepts and
Engineering Applications

Alireza Askarzadeh
Kerman Graduate University of Advanced Technology, Iran

Esmat Rashedi
Kerman Graduate University of Advanced Technology, Iran

ABSTRACT
Harmony search (HS) is a meta-heuristic search algorithm which tries to mimic
the improvisation process of musicians in finding a pleasing harmony. In recent
years, due to some advantages, HS has received a significant attention. HS is easy
to implement, converges quickly to the optimal solution and finds a good enough
solution in a reasonable amount of computational time. The merits of HS algorithm
have led to its application to optimization problems of different engineering areas.
In this chapter, the concepts and performance of HS algorithm are shown and some
engineering applications are reviewed. It is observed that HS has shown promising
performance in solving difficult optimization problems and different versions of this
algorithm have been developed. In the next years, it is expected that HS is applied
to more real optimization problems.

DOI: 10.4018/978-1-5225-2322-2.ch001

Copyright ©2017, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited.
 Harmony Search Algorithm

INTRODUCTION

Most often, engineering optimization problems have non-linear and non-convex ob-
jective functions with intense equality and inequality constraints along with various
types of decision variables. As a result, solving such optimization problems using
traditional methods faces with increasing difficulties. Meta-heuristic optimization
algorithms can be efficient alternatives to conquer the difficulty of complex optimi-
zation problems. Originally invented in (Geem et al. 2001), HS is a meta-heuristic
optimizer inspired by the music improvisation process. In music improvisation
process, a predefined number of musicians attempt to tune the pitch of their instru-
ments to achieve a pleasing harmony (best state). In nature, a harmony is defined
by a special relation between several sound waves that have different frequencies.
The quality of the improvised harmony is determined by aesthetic estimation. In
order to improve the aesthetic estimation and find the best harmony, the musicians
make practice after practice.
There are similarities between musicians improvisation and optimization pro-
cesses. In an optimization problem, the ultimate aim is to find the global optimum
of the objective function under consideration by tuning a predefined number of
decision variables. Indeed, in an optimization problem the decision variables make
a solution vector. Then, the values of the decision variables are put into the objective
function and the quality of the solution vector is calculated. The solution vector is
updated during the iterations until the global optimum is obtained.
Comparison of musical and optimization processes reveals the following simi-
larities:

• In musical process, the quality of a harmony is determined by aesthetic esti-

mation. In an optimization process, the quality of a solution vector is deter-
mined by the objective function value.
• In musical process, the ultimate goal is to obtain the best (fantastic) harmony.
In an optimization process, the ultimate goal is to obtain the global optimum.
• In musical process, musicians change the pitch of their instruments. An opti-
mization algorithm changes the values of the decision variables.
• In musical process, any attempt to play a harmony is called practice. In an
optimization, each attempt to update a solution vector is called iteration.

In general, when a musician wants to tune the pitch of his/her instrument (for
example fiddle, saxophone, etc) and sound a note, he/she utilizes one of the three
possible ways. Theses rules are the main body of HS algorithm.

2
Harmony Search Algorithm

1. He/she can sound a note from the possible range randomly.

2. He/she can sound a note from his/her memory.
3. He/she can sound a note near-by one note from his/her memory.

In the following next sections, HS implementation for optimization, HS variants

and HS applications have been explained in detail.

HS IMPLEMENTATION FOR OPTIMIZATION

In HS algorithm, a feasible solution is called harmony and each decision variable

of the solution is corresponding to a note. HS includes a harmony memory (HM)
in which a predetermined number of harmonies (N) have been stored. Suppose the
goal is to minimize/maximize a fitness function (f) subject to d decision variables.
This optimization problem is defined as follows:

(
Min. or Max. ) f ( x1, x2 ,..., xd ) (1)

where f is the fitness function, xi (i = 1,2,…,d) is decision variable i and d denotes

the problem dimension.
In order to implement HS algorithm for optimization, the following steps should
be used:

Step 1. A harmony memory is initialized.

Step 2. A new harmony is improvised.
Step 3. The new harmony is included in the HM or excluded.
Step 4. Steps 2 and 3 are repeated until the stopping criterion is met. When the
stopping criterion is met, go to
Step 5. The best harmony stored in HM is returned as the found optimum solution.

The details of each step are as follows:

Step 1. Initialization of HM

In HS, at first, N harmonies are produced in the search space and stored in HM.
Table 1 shows the structure of the HM.
As Table 1 shows, harmony i can be specified by a vector, harmony i = [xi,1xi,2
… xi,d]. In order to initialize the HM, Eq. (2) can be used. The last column of HM
is the fitness function values corresponding to each harmony. For example, f1 means

3
 Harmony Search Algorithm

Table 1. The structure of harmony memory

x1 x2 … xd f
Harmony 1 x1,1 x1,2 … x1,d f1
Harmony 2 x2,1 x2,2 … x2,d f2
⁝ ⁝ ⁝ ⁝ ⁝
Harmony N xN,1 xN,2 … xN,d fN

the value of the fitness function for harmony 1. This value is calculated by putting
the decision variables of harmony 1 into the fitness function.

(
xi, j = l j + rand × ui − l j ) i = 1, 2,..., N ; j = 1, 2,..., d (2)

where lj and uj are the lower and upper bounds of decision variable j, respectively,
and rand is a random number with uniform distribution from [0 1]. Mathematically,
the HM is shown by the following expression:

 Harmony 1   x x1,2 … x1,d f1 

   1,1
 Harmony 2   x x2,2 … x2,d f2 
HM =   =  2,1 (3)
      
  
   
 Harmony N   xN ,1 xN ,2 … xN ,d fN 


Step 2. Improvisation of a New Harmony

The next step is to improvise a new harmony, xnew = [xnew,1xnew,2 …xnew,d ]. The main
feature of HS algorithm in comparison with other metaheuristics such as genetic
algorithm (GA) is that in HS, a new harmony is generated by use of all the existing
harmonies. For generation of decision variable j, the following procedure is con-
ducted. This procedure in done for all the decision variables until a new harmony
is obtained.

Stage 1

A random number with uniform distribution from [0 1] is generated (rand). If rand

> HMCR, the decision variable of the new harmony (xnew,j) is randomly generated
by Eq. (4). HMCR which is the abbreviation of harmony memory considering rate
varies between 0 and 1.

4
Harmony Search Algorithm

( )
xnew, j = l j + rand × ui − l j (4)

Otherwise, if rand ≤ HMCR , one of the harmonies stored in HM is randomly

selected, for example k where 1 ≤ k ≤ N . Then, xnew,j is selected by the correspond-
ing value of harmony k from HM as Equation 5.

xnew, j = xk , j (5)

Stage 2

In order to escape from local optima, HS makes use of pitch adjustment mechanism
by which the improvised note may be shifted to a neighbour value with respect to
the possible range. In HS, there is a parameter named pitch adjusting rate (PAR)
which varies between 0 and 1. Small values of PAR leads to having a weak pitch
adjustment mechanism and large values of PAR result in having a rich pitch adjust-
ment mechanism. In order to perform pitch adjustment mechanism, after stage 1, a
random number from [0 1] with uniform distribution is generated (rand). If
rand ≤ PAR , the improvised note should be shifted to a neighbour value by using
Equation 6. Otherwise, if rand > PAR, the improvised note does not change.

xnew, j = xnew, j + bw × (rand − 0.5) × u j − l j (6)

where bw is called the bandwidth of generation and rand is a random number be-
tween 0 and 1 with uniform distribution. The term of (rand − 0.5) generates a
random number from [-0.5 0.5]. Since the users donot know that it is better to in-
crease the value of the decision variable or decrease its value, the term of (rand − 0.5)
is used to randomly select the direction of movement. In Eq. (6), the term of u j − l j
is used to control the scale of the decision variables since in an optimization prob-
lem, the scale of the decision variables may vary significantly such as -104 to 104
in one dimension and -10-5 to 10-5 in another one.

Step 3. Replacement

After Step 2, we have a new feasible harmony. The fitness function of the new
harmony (fnew) is calculated. In this Step, we compare the new harmony and the
worst harmony stored in HM. Suppose harmony h (1 ≤ h ≤ N ) is the worst har-

5
 Harmony Search Algorithm

mony stored in HM in terms of the fitness function value. In this case, if fnew is
better than fith, harmony h is removed from the HM and the new harmony is replaced.
Otherwise, if fnew is worse than fith, the new harmony is abandoned.

Step 4. Stopping Criterion

Most often, the stopping criterion of optimization algorithms is to reach to a pre-

defined number of iterations (tmax). When this criterion is met, the algorithm is
terminated and we go to Step 5. Otherwise, Steps 2 and 3 are repeated until the
stopping criterion is satisfied.

Step 5. Final result

In this Step, the best harmony stored in HM is returned as the optimum solution of
the problem under consideration. Figure 1 shows the flow chart of HS implementa-
tion for optimization.
HS has three parameters, namely, HMCR, PAR and bw. The role of these param-
eters on HS performance is as follows:

• HMCR: The value of HMCR which belongs to [0 1], denotes the probability
of using historical values stored in HM for playing a note. Small values of
HMCR result in random search (with the probability of 1-HMCR) and vice
versa. For example, HMCR of 0.9 means that the note will be sound from the
HM by the probability of 0.9 and will be sound randomly by the probability
of 0.1 from the possible range.
• PAR: By this value which belongs to [0 1], each value improvised from the
HM, has a change to be replaced by a value located at the vicinity of the se-
lected value from HM. In HS, this chance is provided and controlled by the
probability of PAR. As a result, large values of PAR increase the probability
of pitch adjustment and vice versa.
• bw: If the pitch adjustment mechanism is selected, the value of bw controls
the step size of movement. By using large values of bw, the distance between
the new value and the HM value increases. Indeed, we can tune the global and
local search by bw value.

As the first investigation, the Rosenbrock’s banana test function, defined by

Equation 7, is used to illustrate the performance of HS algorithm. Figure 3 indi-
cates the schematic of the test function. The global minimum of this function is 0
which is located at x1 = 1 and x2 = 1. HS algorithm is used to find the solution of

6
Harmony Search Algorithm

the Rosenbrock’s banana test function. The parameter setting of HS is as follows:

N = 20, HMCR = 0.95, PAR = 0.8, bw = 0.2 and tmax = 10000.

Min. (
f ( x1, x2 ) = Ln 1 + (1 − x1 )2 + 100( x2 − x12 )2 ) (7)
−10 ≤ x1, x2 ≤ 10

HS algorithm can successfully find the global solution of the Rosenbrock’s

banana test function. Figure 4 shows the position of the initial solutions stored in
HM. Figures 5 to 8 shows the position of the produced solutions after 100, 1000,
5000 and 10000 iterations, respectively. As can be seen, HS algorithm can provide
a good balance between diversification and intensification during its search process.
In order to evaluate the search power of the classical HS algorithm, Griewank
test function in 30 dimensions is used as a multimodal high-dimensional test func-
tion where there are many local optima over the search range. The expression of
this function is shown by Equation 8. The global minimum of this function is 0
which is located at (0,0,….,0).

1 n n xi
Min. f ( x) =
4000
∑ i=1
( xi )2 − ∏ i=1
cos(
i
)+1
(8)
−600 ≤ xi ≤ 600 i = 1, 2,..., 30

In order to solve the Griewank test function, the parameter setting of HS is as

follows: N = 50, HMCR = 0.9, bw = 0.1 and tmax = 50000. Table 1 shows the impact
of different PAR values on the performance of HS over the Griewank test function.
The statistical results reported in this table have been obtained over 30 independent
runs. As can be seen, the best performance is achieved when the value of PAR is set
to 0.1. In Table 1, PAR = rand means that the value of PAR is adjusted by a random
number between 0 and 1 drawn from a uniform distribution.
Figure 8 and Figure 9 show the convergence rate of HS algorithm during the first
5000 iterations. These figures show the fitness function value of the best harmony
stored in HM at each iteration. It is clear that HS finds a good region of the search
space at the first iterations and tries to converge to the solution.
Table 2 represents the impact of different bw values on the performance of HS
over the Griewank test function. In this case, the parameter setting of HS is as fol-
lows: N = 50, HMCR = 0.9, PAR = 0.1 and tmax = 50000. It is observed that by use
of bw = 0.01, HS can find better results than the other values for the studied test
function.

7
 Harmony Search Algorithm

Figure 1. Flow chart of harmony search algorithm for optimization

8
Harmony Search Algorithm

Figure 2. The graph of the Rosenbrock’s banana test function

Figure 3. The initial positions memorized in HM

9
 Harmony Search Algorithm

Figure 4. The positions generated by HS algorithm during the first 100 iterations

Figure 5. The positions generated by HS algorithm during the first 1000 iterations

10
Harmony Search Algorithm

Figure 6. The positions generated by HS algorithm during the first 5000 iterations

Figure 7. The positions generated by HS algorithm during the first 10000 iterations

11
 Harmony Search Algorithm

Table 2. Performance of HS algorithm on Griewank function considering different

PAR values

Index PAR = 0.1 PAR = 0.3 PAR = 0.5 PAR = 0.7 PAR = 0.9 PAR = rand
Average 1.39 2.86 4.59 6.38 7.89 4.70
Standard 0.07 0.20 0.47 0.44 0.97 0.32
Deviation
Minimum 1.26 2.39 3.28 5.53 4.93 4.07
Maximum 1.54 3.35 5.32 7.12 9.90 5.45

Table 3. Performance of HS algorithm on Griewank function considering different

bw values

Index bw = 0.001 bw = 0.01 bw = 0.1 bw = 1 bw = rand

Average 1.20 1.03 1.39 9.33 3.74
Standard 0.06 0.03 0.07 2.02 0.46
Deviation
Minimum 1.09 0.87 1.26 6.06 2.53
Maximum 1.31 1.07 1.54 13.47 4.39

COMPARATIVE STUDY OF VARIOUS VARIANTS

Investigations on HS performance show that this algorithm has a good ability in

exploration and can discover the potential solution rapidly. However, the local search
ability of HS is weak so that no better solution is expected at the latter iterations of
HS algorithm. In the literature, various variants have been devised to improve the
searching ability of HS.
Like other optimization algorithms, parameter setting of HS algorithm is a chal-
lenging task. Since PAR and bw have a great influence on the quality of the final
solution, proper parameter setting of HS increases the probability of finding the
global solution more and more. The absence of general rules for doing parameter
setting of HS algorithm has guided the research towards developing new variants
of HS which focus on parameter setting.
In this context, one of the most famous and popular variants of HS algorithms is
the investigation made by (Mahdavi et al. 2007). They have proposed time-varying
values for PAR and bw. Based on their suggestion, the value of PAR increases lin-
early during the iterations as follows:

12
Harmony Search Algorithm

Figure 8. Convergence rate of HS algorithm for solving the Griewank function

considering different PAR values during the first 5000 iterations
For a more accurate representation of this figure, please see the electronic version.

Figure 9. Convergence rate of HS algorithm for solving the Griewank function

considering different bw values during the first 5000 iterations
For a more accurate representation of this figure, please see the electronic version.

13
 Harmony Search Algorithm

PARmax − PARmin
PAR t = PARmin + ×t (9)
tmax

where PARt is the PAR value at iteration t, tmax is the maximum number of iterations,
PARmin is the minimum PAR value and PARmax is the maximum PAR value.
The value of bw decreases nonlinearly during the iterations by the following
exponential function:

 bw t 
bwt = bwmax × exp  Ln( min ) ×  (10)
 bwmax tmax 

where bwmax and bwmin are the maximum bandwidth and minimum bandwidth,
respectively.
Based on this idea that a successful search should be proceeded progressively at
the beginning of the algorithm and then gradually settled down, (Wang and Huang
2010) have developed a self-adaptive HS algorithm which utilizes a decreasing linear
PAR during the iterations to prevent overshooting and oscillation. Indeed, this idea
is the opposite of the idea used by (Mahdavi et al. 2007). Also, they have used the
maximal and minimal values in the HM to conduct pitch adjustment mechanism
instead of using the parameter of bw. For this aim, one of the following equations
is used for doing the pitch adjustment:

xnew, j = xnew, j + max( HM j ) − xnew, j  × rand (11)

 

xnew, j = xnew, j −  xnew, j − min( HM j

) × rand (12)
 

where max( HM j ) and min( HM j ) are the highest and the lowest values of vari-
able j in HM. By use of this pitch adjustment mechanism, the decision variables
will not violate the boundary constraint.
There are other investigations which have developed HS variants with the aim
of finding the best parameter setting. In (Geem 2006; Mukhopadhyay et al. 2008),
optimum setting of bw has been discussed. In (Geem et al. 2005; Chakraborty et al.
2009), the parameter of PAR has been modified. In the first, it is proposed to replace
the PAR parameter by a mutation operator borrowed from differential evolution
(DE) algorithm and in the latter, a multi-pitch adjusting rate has been proposed. In
(Hasancebi et al. 2009), the dynamic variation of HMCR and PAR values has been
proposed.

14
Harmony Search Algorithm

There are investigations which have tried to eliminate the burden of manually
finding the best parameter setting of HS algorithm. In (Geem, Sim 2010), an ad-
ditional matrix is used for memorizing operation types (random selection, memory
consideration, or pitch adjustment) in each variable. At each iteration, the parameters
of HMCR and PAR are recalculated based on the matrix information. In (Askar-
zadeh, Zebarjadi 2014), the following equations have been proposed for tuning
the HS parameters based on random numbers drawn from a uniform distribution
between 0 and 1:

HMCR = 0.9 + 0.1 × rand (13)

1 − rand
PAR = (14)
2

bw = rand (15)

In the pitch adjustment mechanism of original HS algorithm, the direction is

selected randomly and the step size is determined by bw. Some researchers have
tried to introduce new pitch adjustment mechanisms. In (Omran, Mahdavi 2008),
a global-best HS (GHS) has been developed in which the pitch adjustment is only
based on the best harmony of HM and there is no need to bw. In (Wang, Huang
2010), a variant, named self-adaptive HS (SAHS), has been proposed in which a
new harmony is improvised based on the maximal and minimal values of HM. In
some investigations, the hybridization of HS algorithm with concepts borrowed from
the other meta-heuristic algorithms such as differential mutation (Wang, Li 2013)
and particle swarm optimization (Pandi, Panigrahi 2011) has been investigated and
satisfactory results have been reported.

APPLICATIONS

To date, HS is utilized to solve various engineering optimization problems. Some

of the works have been reviewed below.

Electrical Engineering

The hybridization of HS and DE was developed in (De et al. 2015) for designing a
CMOS inverter. The heuristic algorithm found the best design variables to optimize
the system performance. The design variables were the ratio of the channel width
to the channel length of the CMOS transistors and the output load capacitance.

15
 Harmony Search Algorithm

Telecommunication

In telecommunication, HS was used for optimal designing of wireless sensor net-

works (WSN), antenna and radar. The pattern synthesis of linear antenna arrays was
investigated in (Guney,Onay 2011). Pattern synthesis is the process of determining
the parameters of an antenna array to achieve the target antenna radiation pattern.
HS with a local search procedure utilized to solve the Spread-Spectrum Radar
Polyphase (SSRP) codes design in (Gil-López et al. 2012). The variables are the
magnitudes of the phase differences for the SSRP and the goal was the minimization
of the module of the largest among the samples of the autocorrelation function. HS
with a local search procedure was used for node localization in the wireless sensor
networks with noisy distance-related measurements in (Manjarres et al. 2013). There
were two objective functions of CF and CV. CF is the squared error between the
estimated and the measured inter-node distances of the nodes that are in the con-
nectivity range of each others and CV is the number of connectivity neighborhood
constraints which are not satisfied by the candidate topology. The design of low
complexity sharp Modified Discrete Fourier Transform (MDFT) filter bank using
hybrid HS and Gravitational search algorithm were reported in (Sakthivel,Elias
2015). Filter coefficients are optimized to design a low complexity sharp MDFT
filter bank with low power consumption, low chip area and high speed of operation.

Pattern Recognition and Image/Speech Processing

In the field of pattern recognition, there are two famous topics of classification and
clustering. HS could be adopted to design a classifier or a clustering algorithm.
The application of HS in recognizing patterns in image or speech data are reported
in some researches.

Classification

Classification is the process of assigning category labels to any type of data. Two
important parts of classification systems are classifiers and features. HS could help
to improve the accuracy of the system by improving the classifier or selecting the
best set of features. In (Kulluk et al. 2012), six benchmark classification problems
were solved by an artificial neural network (ANN) which trained using self-adaptive
global best version of HS. The process of training feed-forward NNs using HS was
performed by optimizing the weights of links between the neurons to minimize the
error. A hybrid harmony-based classifier was developed in (Karimi et al. 2012). HS
produced if-then rules that determine the label of the data according to its attributes.
Authors in (Wang et al. 2015b) used HS as a feature selection method for email

16
Harmony Search Algorithm

classification. HS found the best discriminative features in regard to the document

frequency and term frequency. Email classification could help for filtering spam.
A self-adaptive HS utilized in (Huang et al. 2014) for local feature selection in the
classification of music genre. Music features were five acoustic characteristics and
the number of music genres was ten. The goal was the maximization of classification
precision. In (Inbarani et al. 2015), the hybridization of HS and rough set theory was
developed to select the best set of features to maximize the cassification accuracy. A
hybrid HS combined with stochastic local search was produced in (Nekkaa,Boughaci
2015) to find the beneficial features for the classification outcome optimization.
The support vector machine (SVM) was used as the classifier and its parameters
(penalty parameter C and the gamma parameter for the RBF kernel) were tuned by
an iterative search. A self-adjusting HS was produced in (Zheng et al. 2015a) to
maximize the classification accuracy of C4.5 classifier by feature selection.

Clustering

Clustering is about discriminating data into some groups (called clusters), in such
a way that data in a group are similar to each other and different from the other
groups. Harmony K-means algorithm was produced in (Mahdavi,Abolhassani 2009)
for document clustering. Documents were represented using term weights. HS found
the center of the clusters and then the results were refined by K-means.

Image Processing

In the image processing algorithms, heuristic optimization algorithms like HS are

useful for optimizing, tuning and enhancing the processing results. There are various
works in this regard. In (Cuevas 2013), block matching for motion estimation was
handeled by HS. The fitness function was the matching quality of each motion vector
candidate. A window arround the current block was searched for the best matched
block. The motion vector is the position difference between the current block and
the best match block. In images, the saliency map is calculated by combining vari-
ous information like color, intensity, orientation, and other feature maps. Saliency
map is useful for target detection. To have more target conspicuity in the saliency
map, a modified Gaussian HS was used in (Li,Duan 2014). Tuning the parameters
of Deep Belief Networks was performed using HS in (Papa et al. 2015). Binary
mage reconstruction was solved by this network to minimize the mean square error
beween reconsrtucted and original image. Image restoration by projections onto
convex sets was tackled in (Pires et al. 2015) where its parameters were optimized
by HS. HS found the best parameters to maximize the improvement signal to noise
ratio (ISNR).

17
 Harmony Search Algorithm

Speech Processing

HS was used to solve part-of-speech (PoS) tagging problem in (Forsati,Shamsfard

2015) and find the best tagging quality. PoS is about syntactic tagging to every word
in a sentence according to the context.

Water Management

To design and manage water distribution and water supply networks, there are some
optimization problems that could be handled by HS. Water distribution network
design problem was solved in (Geem 2012) by a diversity improved version of HS
to minimize the construction cost. The problems have design constraints of nodal
pressure and quantity and the decision variables are the diameter of the pipes. In
(Baek et al. 2010), HS was employed to optimize the simulation of hydraulic under
abnormal operating conditions in water distribution systems. HS built the model to
minimize the error between the assumed and the calculated heads at the demand
nodes. Pump scheduling problem in water distribution systems was tackled in
(Kougias,Theodossiou 2013) using the polyphonic HS. The objectives were the
minimization of quantity of pumped water and electricity and maintenance cost. The
design of groundwater remediation systems was tackled by a probabilistic multi-
objective fast HS in (Luo et al. 2014; Luo et al. 2012) to minimize the remediation
cost and contaminant mass remaining in aquifer. Authors in (Atrabi et al. 2015) used
HS for reservoir operation optimization to minimize the water supply deficit and
flood damages of a reservoir. The subjected constraint was mass balance limitation.

Mechanical Engineering

In mechanical engineering, HS was utilized for optimizing mechanical process

and mechanical design problems. Authors in (Abhishek et al. 2016) used a fuzzy
embedded HS for tuning the machining process parameters of the carbon fiber
reinforced polymer (CFRP) composites. These composites are used for manufactur-
ing engineering components especially in aerospace and automobile industries. A
fuzzy inference system was used to model the multiple performance characteristics
into a single objective function. Some well-known constraint mechanical design
optimization problems namely the pressure vessel design problem, the tension/
compression spring design problem, and the welded beam design problem were
solved in (Mun,Cho 2012) by a modified HS. Authors in (Razfar et al. 2011) found
optimal cutting parameters in face milling process by HS to minimize the surface
roughness. The surface roughness was modeled by feed forward artificial neural

18
Harmony Search Algorithm

networks. Surface roughness formation and cutting force mechanisms modeling

were also developed in (Zinati,Razfar 2012) by HS and artificial neural networks.

Chemical Engineering

Thermodynamic models in chemical engineering are useful for synthesis, design,

optimization, and control of process systems (Merzougui et al. 2012). The param-
eter estimation of these models could be solved by heuristic optimization. HS was
applied in (Merzougui et al. 2012) for estimation of binary interaction parameters
in the liquid–liquid phase equilibrium model. The objective function was the data
fitting between observed and calculated data. The support vector regression (SVR)
estimated the relationship function between amount of asphaltene precipitation and
titration data in (Fattahi et al. 2015). HS was tuned the SVR learning parameters and
the data fit accuracy was defined as the objective function. This method is useful
in petroleum industry.

Power Engineering

In the field of power engineering, there are many optimization problems like eco-
nomic dispatch (ED), optimal power flow (OPF), unit commitment (UC), modeling
problems, designing, and power system planning problems. HS was examined to
solve these problems in various reported works.

Economic Dispatch

ED is defined by optimal determination of the generator power outputs while support-

ing the load demand. It has two types of static or dynamic dispatch and it has some
equality and inequality constraints. Authors in (Arul et al. 2013a) utilized HS with
chaotic self-adaptive differential mutation operator for optimal solving of dynamic
ED with some constraints. The constraints were about real power balance, real power
generation limits, and generating unit ramp-rate limits. The objective function was
the total fuel cost. Combined heat and power economic dispatch optimized by HS in
(Javadi et al. 2012). Three types of thermal units including co-generation, electrical
only, and heat-only units were considered and the energy production cost was mini-
mized. Authors in (Jeddi,Vahidinasab 2014) employed modified HS for economic
load dispatch to minimize the generation cost while satisfying unit operating limits,
power balance constraints, ramp-rate limits, and spinning reserve constraints. The
modified HS was produced using wavelet mutation and a memory consideration
scheme based on the roulette wheel. Authors in (Khazali,Kalantar 2011) used HS
to find the settings of control variables to optimize the power transmission loss,

19
 Harmony Search Algorithm

voltage stability, and voltage profile in economic power dispatch problem. Control
variables were generator voltages, tap positions of tap changing transformers and the
amount of reactive compensation devices. Authors in (Niu et al. 2014) proposed HS
with arithmetic crossover operation for solving five different types of ED problems,
including static dispatch with valve point effects, ED with prohibited operating
zones, ED with multiple fuel cells, combined heat and power ED, and dynamic ED.
A differential HS was employed in (Wang,Li 2013) for solving the non-convex ED
to minimize the total fuel cost. Constraints of generating capacity, power balance,
ramp rate limit, and prohibited operating zones were handled by a repair procedure
and three simple selection rules. Authors in (Arul et al. 2013a) solved non-convex
ED using HS for generation cost minimization. The constraints of real power bal-
ance, generation limit, prohibited operating zones, ramp-rate, and system spinning
reserve, were tackled by penalization. Non-convex ED was also solved in (Arul et
al. 2014) by an improved version of HS.

Optimal Power Flow

OPF is defined by the minimization of generation costs and loss in power system
variables under some equality and inequality constraints. OPF problem was solved
using improved HS in (Sinsuphan et al. 2013) to minimize the total production cost
under constraints for real and reactive power flow and variable limits. OPF was
also solved by HS in (Sivasubramani,Swarup 2011) with the objectives of the total
fuel cost, real power loss, and voltage stability index minimization. The authors in
(Arul et al., 2013b) proposed a chaotic self-adaptive differential HS to solve OPF
problems with non-smooth and non-convex cost functions.

Unit Commitment

UC is the problem of determining schedules of generating units including on/off

states of the units and how much they generate hourly. The objective of UC is to
minimize the total operating cost. UC is a large-scale, non-convex, and constraint
problem. Security constrained unit commitment problem was handled in (Samiee
et al. 2013) to minimize the fuel cost and startup and shut down costs of thermal
units while satisfying constraints for: AC power flow constraints, up/down time
duration, ramp rate, operating zones, system spinning, AC security, and reactive
power generation. To handle these constraints, a hybrid method composed of tri-
state algorithm and infeasible solution rejection technique was produced. The unit
commitment problem was solved by an improved version of HS in (Morsali et al.
2014) to minimize the operation, start-up and shut-down costs. This objective func-
tion was subjected to the load demand constraints and ramp rate limits. To solve this

20
Harmony Search Algorithm

problem, HS control parameters of memory size, HMCR, PAR and bw was adapted.
In (Paqaleh et al. 2010), to solve UC by HS, the considered objectives were the fuel
costs of generating units, the start-up costs of the committed units and shut-down
costs of the decommitted units and the constraints were the power balance, spinning
reserve capacity of generating units, unit ramp-up and ramp-down rates, minimum
up/down time limit, and spinning reserve requirement.

Modeling

Grouping-based global HS was proposed in (Askarzadeh,Rezazadeh 2011) for

modeling the voltage as a function of the current in proton exchange membrane
fuel cell (PEMFC). HS found the parameters of the modele to fit the simulated
data and minimize the error. Authors in (Askarzadeh,Zebarjadi 2014) used HS for
wind power modeling by finding the optimum parameters of the model. Optimum
parameters minimize the error between the model and actual wind powers. Authors
in (Askarzadeh,Rezazadeh 2012) utilized HS for modeling the current as a function
of the voltage in solar cells. HS identified the model parameters to minimize the
difference between experimental data and the model results.

Designing and Planning

Optimum design of a PV/wind hybrid system was experimented in (Askarzadeh

2013b). A hybridization of HS with chaotic search and simulated annealing was
used to find the number of PV panels, wind turbines and batteries which minimize
the total annual cost of the system in the presence of some constraints. A discrete
HS was proposed in (Askarzadeh 2013a) for determining the optimum size of wind–
photovoltaic hybrid energy systems which minimize the total annual cost. Power
network partitioning was solved by HS in (Ezhilarasi,Swarup 2012b) to minimize
the number of nodes in a cluster and the interconnections between the clusters.
Optimally partitioning the network into clusters enhances the computational com-
plexity by parallel processing. This problem was also solved in (Ezhilarasi,Swarup
2012a) by HS and a graph bi-partitioning method called Kernighan–Lin strategy.
Authors in (Javaheri,Goldoost-Soloot 2012) employed HS for optimal locating and
sizing of FACTS devices in power systems to minimize the total generation cost.
The optimal placement and sizing of capacitors in power distribution networks
was handled in (Sirjani,Bade 2015) by an improved global HS, to minimize the
power loss and total cost. The problem constraints were load unbalancing, mutual
coupling and harmonics. Power transmission expansion planning was solved in
(Rastgou,Moshtagh 2014) by improved HS to minimize the construction costs,
congestion cost, and security cost. A hybrid self- adaptive global HA was utilized

21
 Harmony Search Algorithm

in (Shivaie et al. 2015) for reliability-based Distribution Expansion Planning (DEP).

There were four objectives of the investment, maintenance, operation, and expected
customer interruption costs with the problem constraints of operational restrictions,
Kirchhoff’s laws, radial structure limitation, voltage limits, and capital expenditure
budget restriction.

Civil Engineering

In civil engineering, there are many complicated optimization problems that was
solved by HS such as optimizing building structures, road construction, and solving
structural design problems.

Optimizing Structural Designs

Special seismic moment reinforced concrete frames are used in buildings to resist
earthquakes. Ref (Akin,Saka 2015) utilized HS for optimal design of these frames
under earthquake loads to minimize the frame cost. The frame design variables
are consisted of two groups of column design variables and beam design vari-
ables. Authors in (Amini,Ghaderi 2013) used the hybridization of HS with ant
colony search algorithm for optimal locating of dampers in structural systems.
The objective function was defined by the dynamic response of the system and the
configuration and forces of the dampers which was minimized. Two improved HS
algorithms were proposed in (Degertekin 2012) for optimally designing of truss
structures. Size and shape of truss structures was also optimized in (Kaveh,Javadi
2014) with a hybrid method of PSO and HS to minimize the truss weight. A social
HS model was employed for design of the composite floor with minimum cost in
(Kaveh,Ahangaran 2012). Design variables were related to the concrete slab, steel
beams, and the shear studs. Four classical weight minimization problems of steel
frames were designed using an enhanced HS in (Maheri,Narimani 2014). Damage
under ambient vibration in structural systems was detected by HS in (Miguel et al.
2012). HS produced numerical model for structural damage and minimized its dif-
ference with the experimental model. The structural damage was estimated from a
model update process using damage-induced changes in the modal features. Steel
frame optimization was solved in (Murren,Khandelwal 2014) to obtain least-weight
designs subjected to design and drift constraints. A hybrid HS was proposed in
(Segura et al. 2015) for optimizing the sustainability of post-tensioned concrete box-
girder pedestrian bridges. HS found the structural design variables and the geometry
for minimizing the costs while satisfying the constraints for structural safety and
durability. Authors in (Bekdas 2015) designed post-tensioned axially symmetric
cylindrical reinforced concrete walls and minimized the total material cost using

22
Harmony Search Algorithm

HS. These walls are used in storage tanks for liquids and solids. In (Carbas,Saka
2012), topological design of domes was performed by improved HS to minimize the
weight of each dome. Optimal design of steel frames was developed in (Degertekin
2008) by HS to determine the design variables of steel and minimize the weight of
the frame. Optimal sensor placement on gantry crane structures was solved by HS
in (Jin et al. 2015). These sensors are used to evaluate the safety and reliability of
structures. The number and the location of the sensors were determined by HS to
improve the performance. Optimal design of geosynthetic-reinforced retaining walls
was performed in (Manahiloh et al. 2015) by HS to minimize the construction cost
in the presence of design constraints.

Road Transportation

The optimum signal settings in a road transportation with a hybridization of HS

and hill climbing was performed in (Ceylan,Ceylan 2012) for minimizing the travel
time and the fuel consumption. In (Salcedo-Sanz et al. 2013), the reconfiguration of
one-way roads in a city by providing alternative routes was solved by a two-objective
HS. The proposed method guaranteed the mobility of citizens. The value of this
function was estimated by a Monte Carlo simulation. A modified HS was produced
in (Lee,Mun 2014) to establish a dynamic model for describing the resistance for
rutting and fatigue cracking of asphalt concrete mixture. HS found the parameters
of the model to fit to the laboratory tests.

Financial Tasks

A self adaptive differential HS was used in (Dash et al. 2014) for designing a single
hidden layer feed forward neural network. The optimized neural network was used
for prediction of a financial time series data that was the closing price and volatility
of five different stock indices. Forecasting of financial data was performed in (Dash
et al. 2015) by an integrating model of a interval type2 fuzzy logic system and an
artificial neural network. A proposed differential HS was used for optimizing the
parameters of the fuzzy time series models for stock market volatility prediction.
A stochastic replenishment intervals multiproduct inventory model with dynamic
demand was solved by HS in (Taleizadeh et al. 2012) to minimize costs of holding,
purchasing and shortage. Multi-site order planning problem was handled by HS
in (Guo et al. 2015). This problem was defined in a make-to-order manufacturing
environment considering multiple production uncertainties. The goal is to attain
production objectives by effectively determining the allocation of customer orders
to several self-owned or collaborative production plants placed in different regions.

23
 Harmony Search Algorithm

Product Transportation

Multi-server location–allocation problem was tackled by multi-objective HS in (Ha-

jipour et al. 2014). This problem is about placing a number of facilities in between
a number of customers located at fixed points to minimize the transportation cost.
Shipping cost in a transportation problem was minimized by hybridization of HS
and simulated annealing in (Hosseini et al. 2014). In this problem, some vehicles
transport goods from suppliers to the customers via three transportation systems.
The problem of emergency air transportation was solved in (Zheng et al. 2015b) by
a hybridization of biogeography-based optimization and HS, to maximize delivery
efficiency. This problem is about the transportation of amounts of n different types
of relief supplies from a set of m air freight hubs to the closest airport to the target
disaster area.

Job Scheduling

Multi-objective flexible job shop scheduling problem solved by a Pareto version

of HS with multiple harmony generation strategies in (Gao et al. 2014). HS with
one-point crossover and iterative local search was employed in (Li et al. 2015) to
solve the flow line manufacturing cell scheduling problem and minimize the total
tardiness and mean total flow time. Cellular manufacturing system is a production
system which needs cell scheduling before starting the production process. Two-
sided assembly lines are kind of assembly lines in which tasks operations can be
performed in them. Non dominated HS was proposed in (Purnomo,Wee 2014) to
solve the task assignment problem in a two-sided assembly line with zone constraints.
The objective was the maximization of the production rate and the distribution of
the workload in the assembly line. Flexible job-shop scheduling problem was solved
in (Gao et al. 2015) by discrete HS. The objectives were the weighted combination
of the maximum of the completion time and the mean of earliness and tardiness.

Function Optimization

There are some enhanced versions of HS for numerical optimization examined by

some standard function optimization: improved HS (Ashrafi,Dariane 2013), HS
with a novel selection schemes (Al-Betar et al. 2012; Al-Betar et al. 2013), island-
based HS (Al-Betar et al. 2015), Geometric Selective HS (Castelli et al. 2014), HS
with dynamic control parameters (Chen et al. 2012), a learning automata-based
HS (Enayatifar et al. 2013), HS with mutation operators (Hasan et al. 2014), global

24
Harmony Search Algorithm

dynamic HS (Khalili et al. 2014), improved HS (Contreras et al. 2014), parameter

adaptive HS (Kumar et al. 2014), intelligent global HS (Valian et al. 2014), HS
with Gaussian mutation (Dai et al. 2015), HS with the population based incremen-
tal learning (Gao et al. 2012), hybridization of HS and cuckoo search (Wang et al.
2015a). High-dimensional multimodal optimization problems was solved by HS in
(Tuo et al. 2015). 0–1 knapsack problems was solved in (Kong et al. 2015a, b) by
new versions of binary HS. Furthermore, it was solved in (Layeb 2013) by a hybrid
quantum HS.

CONCLUSION

As one of the powerful optimization algorithms, HS has attracted significant atten-

tion for solving different types of optimization problems during the recent years.
The advantages of HS algorithm are easy implementation, simple concept, fast
convergence speed and few parameters to adjust. Owing to these merits, HS has been
considerably used to solve various problems in different fields like power system,
communication, software, civil, water engineering, and pattern recognition. Study of
the literature indicates that HS can effectively and efficiently solve different types of
engineering optimization problems. In the literature, various variants of HS algorithm
can be found which focus on improving improvisation process, harmony memory
consideration and parameter setting. There are other variants which try to enhance
the performance of HS by borrowing some ideas from the other metaheuristic such
as GA, PSO, DE and chaotic search. By comparing the results obtained by HS and
other algorithms it can be drawn that HS could be a good candidate to solve complex
optimization problems.

ACKNOWLEDGMENT

The authors would like to thank Editor and Reviewers for their helpful comments.

REFERENCES

Abhishek, K., Datta, S., & Mahapatra, S. S. (2015). Multi-objective optimization

in drilling of CFRP (polyester) composites: Application of a fuzzy embedded har-
mony search (HS) algorithm. Measurement, 77, 222–239. doi:10.1016/j.measure-
ment.2015.09.015

25
 Harmony Search Algorithm

Akin, A., & Saka, M. P. (2015). Harmony search algorithm based optimum detailed
design of reinforced concrete plane frames subject to ACI 31805 provisions. Com-
puters & Structures, 147, 79–95. doi:10.1016/j.compstruc.2014.10.003
Al-Betar, M. A., Awadallah, M. A., Khader, A. T., & Abdalkareem, Z. A. (2015).
Island-based harmony search for optimization problems. Expert Systems with Ap-
plications, 42(4), 2026–2035. doi:10.1016/j.eswa.2014.10.008
Al-Betar, M. A., Doush, I. A., Khader, A. T., & Awadallah, M. A. (2012). Novel
selection schemes for harmony search. Applied Mathematics and Computation,
218(10), 6095–6117. doi:10.1016/j.amc.2011.11.095
Al-Betar, M. A., Khader, A. T., Geem, Z. W., Doush, I. A., & Awadallah, M. A.
(2013). An analysis of selection methods in memory consideration for harmony
search. Applied Mathematics and Computation, 219(22), 10753–10767. doi:10.1016/j.
amc.2013.04.053
Amini, F., & Ghaderi, P. (2013). Hybridization of Harmony Search and Ant Colony
Optimization for optimal locating of structural dampers. Applied Soft Computing,
13(5), 2272–2280. doi:10.1016/j.asoc.2013.02.001
Arul, R., Ravi, G., & Velusami, S. (2013). Solving optimal power flow problems
using chaotic self adaptive differential harmony search algorithm. Electric Power
Components and Systems, 41(8), 782–805. doi:10.1080/15325008.2013.769033
Arul, R., Ravi, G., & Velusami, S. (2013a). Chaotic self-adaptive differential har-
mony search algorithm based dynamic economic dispatch. International Journal of
Electrical Power & Energy Systems, 50, 85–96. doi:10.1016/j.ijepes.2013.02.017
Arul, R., Ravi, G., & Velusami, S. (2013b). Non-convex economic dispatch with
heuristic load patterns, valve point loading effect, prohibited operating zones,
ramp-rate limits and spinning reserve constraints using harmony search algorithm.
Electrical Engineering, 95(1), 53–61. doi:10.1007/s00202-012-0241-y
Arul, R., Ravi, G., & Velusami, S. (2014). An improved harmony search algorithm
to solve economic load dispatch problems with generator constraints. Electrical
Engineering, 96(1), 55–63. doi:10.1007/s00202-012-0276-0
Ashrafi, S. M., & Dariane, A. B. (2013). Performance evaluation of an improved
harmony search algorithm for numerical optimization: Melody Search (MS). En-
gineering Applications of Artificial Intelligence, 26(4), 1301–1321. doi:10.1016/j.
engappai.2012.08.005

26
Harmony Search Algorithm

Askarzadeh, A. (2013a). Developing a discrete harmony search algorithm for size

optimization of wind–photovoltaic hybrid energy system. Solar Energy, 98(Part C),
190–195. doi:10.1016/j.solener.2013.10.008
Askarzadeh, A. (2013b). A discrete chaotic harmony search-based simulated an-
nealing algorithm for optimum design of PV/wind hybrid system. Solar Energy,
97, 93–101. doi:10.1016/j.solener.2013.08.014
Askarzadeh, A., & Rezazadeh, A. (2011). A grouping-based global harmony search
algorithm for modeling of proton exchange membrane fuel cell. International Jour-
nal of Hydrogen Energy, 36(8), 5047–5053. doi:10.1016/j.ijhydene.2011.01.070
Askarzadeh, A., & Rezazadeh, A. (2012). Parameter identification for solar cell
models using harmony search-based algorithms. Solar Energy, 86(11), 3241–3249.
doi:10.1016/j.solener.2012.08.018
Askarzadeh, A., & Zebarjadi, M. (2014). Wind power modeling using harmony
search with a novel parameter setting approach. Journal of Wind Engineering and
Industrial Aerodynamics, 135, 70–75. doi:10.1016/j.jweia.2014.10.012
Atrabi, H. B., Kourosh, Q., Rheinheimer, D. E., & Sharifi, E. (2015). Application
of harmony search algorithm to reservoir operation optimization. Water Resources
Management, 29(15), 5729–5748. doi:10.1007/s11269-015-1143-3
Baek, C. W., Jun, H. D., & Kim, J. H. (2010). Development of a PDA model for
water distribution systems using harmony search algorithm. KSCE Journal of Civil
Engineering, 14(4), 613–625. doi:10.1007/s12205-010-0613-7
Bekdas, G. (2015). Harmony search algorithm approach for optimum design of
post-tensioned axially symmetric cylindrical reinforced concrete walls. Journal
of Optim Theory Application, 164(1), 342–358. doi:10.1007/s10957-014-0562-2
Carbas, S., & Saka, M. P. (2012). Optimum topology design of various geometrically
nonlinear latticed domes using improved harmony search method. Struct Multidisc
Optim, 45(3), 377–399. doi:10.1007/s00158-011-0675-2
Castelli, M., Silva, S., Manzoni, L., & Vanneschi, L. (2014). Geometric Selective
Harmony Search. Information Sciences, 279, 468–482. doi:10.1016/j.ins.2014.04.001
Ceylan, H., & Ceylan, H. (2012). A Hybrid Harmony Search and TRANSYT hill
climbing algorithm for signalized stochastic equilibrium transportation networks.
Transportation Research Part C, Emerging Technologies, 25, 152–167. doi:10.1016/j.
trc.2012.05.007

27
 Harmony Search Algorithm

Chakraborty, P., Roy, G. G., Das, S., Jain, D., & Abraham, A. (2009). An improved
harmony search algorithm with differential mutation operator. Fundamenta Infor-
maticae, 95, 1–26.
Chen, J., Pan, Q., & Li, J. (2012). Harmony search algorithm with dynamic control
parameters. Applied Mathematics and Computation, 219(2), 592–604. doi:10.1016/j.
amc.2012.06.048
Contreras, J., Amaya, I., & Correa, R. (2014). An improved variant of the conven-
tional Harmony Search algorithm. Applied Mathematics and Computation, 227,
821–830. doi:10.1016/j.amc.2013.11.050
Cuevas, E. (2013). Block-matching algorithm based on harmony search optimiza-
tion for motion estimation. Applied Intelligence, 39(1), 165–183. doi:10.1007/
s10489-012-0403-7
Dai, X., Yuan, X., Wu, L. (in press). A novel harmony search algorithm with gauss-
ian mutation for multi-objective optimization. Soft Comput.
Dash, R., Dash, P. K., & Bisoi, R. (2014). A self adaptive differential harmony
search based optimized extreme learning machine for financial time series prediction.
Swarm and Evolutionary Computation, 19, 25–42. doi:10.1016/j.swevo.2014.07.003
Dash, R., Dash, P. K., & Bisoi, R. (2015). A differential harmony search based
hybrid interval type2 fuzzy EGARCH model for stock market volatility predic-
tion. International Journal of Approximate Reasoning, 59, 81–104. doi:10.1016/j.
ijar.2015.02.001
De, B. P., Kar, R., Mandal, D., & Ghoshal, S. P. (2016). Optimal design of high speed
symmetric switching CMOS inverter using hybrid harmony search with differential
evolution. Soft Computing, 20(9), 3699–3717. doi:10.1007/s00500-015-1731-4
Degertekin, S. O. (2012). Improved harmony search algorithms for sizing optimiza-
tion of truss structures. Computers & Structures, 92–93, 229–241. doi:10.1016/j.
compstruc.2011.10.022
Degertekin, S. O., & Hayalioglu, M. S. (2010). Harmony search algorithm for mini-
mum cost design of steel frames with semi-rigid connections and column bases.
Struct Multidisc Optim, 42(5), 755–768. doi:10.1007/s00158-010-0533-7
Enayatifar, R., Yousefi, M., Abdullah, A. H., & Darus, A. N. (2013). A novel harmony
search algorithm based on learning automata. Communications in Nonlinear Science
and Numerical Simulation, 18(12), 3481–3497. doi:10.1016/j.cnsns.2013.04.028

28
Harmony Search Algorithm

Ezhilarasi, G. A., & Swarup, K. S. (2012a). Network decomposition using Ker-

nighan–Lin strategy aided harmony search algorithm. Swarm and Evolutionary
Computation, 7, 1–6. doi:10.1016/j.swevo.2012.07.002
Ezhilarasi, G. A., & Swarup, K. S. (2012b). Network partitioning using harmony
search and equivalencing for distributed computing. Journal of Parallel and Dis-
tributed Computing, 72(8), 936–943. doi:10.1016/j.jpdc.2012.04.006
Fattahi, H., Gholami, A., Amiribakhtiar, M. S., & Moradi, S. (2015). Estimation of
asphaltene precipitation from titration data: A hybrid support vector regression with
harmony search. Neural Computing & Applications, 26(4), 789–798. doi:10.1007/
s00521-014-1766-y
Forsati, R., & Shamsfard, M. (2015). Novel harmony search-based algorithms
for part-of-speech tagging. Knowledge and Information Systems, 42(3), 709–736.
doi:10.1007/s10115-013-0719-6
Gao, K. Z., Suganthan, P. N., Pan, Q. K., Chua, T. J., Cai, T. X., & Chong, C. S.
(2014). Pareto-based grouping discrete harmony search algorithm for multi-objective
flexible job shop scheduling. Information Sciences, 289, 76–90. doi:10.1016/j.
ins.2014.07.039
Gao, K. Z., Suganthan, P. N., Pan, Q. K., Chua, T. J., Cai, T. X., & Chong, C. S.
(2015). Discrete harmony search algorithm for flexible job shop scheduling problem
with multiple objectives. Journal of Intelligent Manufacturing, 27(2), 363–374.
doi:10.1007/s10845-014-0869-8
Gao, X. Z., Wang, X., Jokinen, T., Ovaska, S. J., Arkkio, A., & Zenger, K. (2012).
A hybrid PBIL-based harmony search method. Neural Computing & Applications,
21(5), 1071–1083. doi:10.1007/s00521-011-0675-6
Geem, Z. (2006). Improved harmony search from ensemble of music players. In
Knowledge-based intelligent information and engineering systems (pp. 86–93).
Heidelberg: Springer. doi:10.1007/11892960_11
Geem, Z. W. (2012). Effects of initial memory and identical harmony in global
optimization using harmony search algorithm. Applied Mathematics and Computa-
tion, 218(22), 11337–11343. doi:10.1016/j.amc.2012.04.070
Geem, Z. W., & Sim, K. B. (2010). Parameter-setting-free harmony search algo-
rithm. Applied Mathematics and Computation, 217(8), 3881–3889. doi:10.1016/j.
amc.2010.09.049

29
 Harmony Search Algorithm

Geem, Z. W., Tseng, C. L., & Park, Y. (2005). Harmony search for generalized
orienteering problem: best touring in China. In Advances in natural computation
(pp. 741–750). Berlin: Springer.
Gil-López, S., Ser, J. D., Salcedo-Sanz, S., Pérez-Bellido, Á. M., Cabero, J. M., &
Portilla-Figueras, J. A. (2012). A hybrid harmony search algorithm for the spread
spectrum radar polyphase codes design problem. Expert Systems with Applications,
39(12), 11089–11093. doi:10.1016/j.eswa.2012.03.063
Guney, K., & Onay, M. (2011). Optimal synthesis of linear antenna arrays using a
harmony search algorithm. Expert Systems with Applications, 38(12), 15455–15462.
doi:10.1016/j.eswa.2011.06.015
Guo, Z. X., Yang, C., Wang, W., & Yang, J. (2015). Harmony search-based multi-
objective optimization model for multi-site order planning with multiple uncertainties
and learning effects. Computers & Industrial Engineering, 83, 74–90. doi:10.1016/j.
cie.2015.01.023
Hajipour, V., Rahmati, S. H. A., Pasandideh, S. H. R., & Niaki, S. T. A. (2014).
A multi-objective harmony search algorithm to optimize multi-server location–al-
location problem in congested systems. Computers & Industrial Engineering, 72,
187–197. doi:10.1016/j.cie.2014.03.018
Hasan, B. H. F., Abu Doush, I., Al Maghayreh, E., Alkhateeb, F., & Hamdan, M.
(2014). Hybridizing Harmony Search algorithm with different mutation operators
for continuous problems. Applied Mathematics and Computation, 232, 1166–1182.
doi:10.1016/j.amc.2013.12.139
Hasancebi, O., Erdal, F., & Saka, M. P. (2009). An adaptive harmony search method
for structural optimization. Journal of Structural Engineering, 137, 419–431.
Hosseini, S. D., Akbarpour Shirazi, M., & Karimi, B. (2014). Cross-docking and
milk run logistics in a consolidation network: A hybrid of harmony search and
simulated annealing approach. Journal of Manufacturing Systems, 33(4), 567–577.
doi:10.1016/j.jmsy.2014.05.004
Huang, Y.-F., Lin, S.-M., Wu, H.-Y., & Li, Y.-S. (2014). Music genre classification
based on local feature selection using a self-adaptive harmony search algorithm.
Data & Knowledge Engineering, 92, 60–76. doi:10.1016/j.datak.2014.07.005
Inbarani, H. H., Bagyamathi, M., & Azar, A. T. (2015). A novel hybrid feature selec-
tion method based on rough set and improved harmony search. Neural Computing
& Applications, 26(8), 1859–1880. doi:10.1007/s00521-015-1840-0

30
Harmony Search Algorithm

Javadi, M. S., Esmaeel Nezhad, A., & Sabramooz, S. (2012). Economic heat and
power dispatch in modern power system harmony search algorithm versus analyti-
cal solution. Scientia Iranica, 19(6), 1820–1828. doi:10.1016/j.scient.2012.10.033
Javaheri, H., & Goldoost-Soloot, R. (2012). Locating and Sizing of Series FACTS
Devices Using Line Outage Sensitivity Factors and Harmony Search Algorithm.
Energy Procedia, 14, 1445–1450. doi:10.1016/j.egypro.2011.12.1115
Jeddi, B., & Vahidinasab, V. (2014). A modified harmony search method for envi-
ronmental/economic load dispatch of real-world power systems. Energy Conversion
and Management, 78, 661–675. doi:10.1016/j.enconman.2013.11.027
Jin, H., Xia, J., Wang, Y.-q. (2015). Optimal sensor placement for space modal
identification of crane structures based on an improved harmony search algorithm.
J Zhejiang Univ-Sci A (Appl Phys & Eng), 16(6), 464-477.
Karimi, Z., Abolhassani, H., & Beigy, H. (2012). A new method of mining data
streams using harmony search. Journal of Intelligent Information Systems, 39(2),
491–511. doi:10.1007/s10844-012-0199-2
Kaveh, A., & Ahangaran, M. (2012). Discrete cost optimization of composite floor
system using social harmony search model. Applied Soft Computing, 12(1), 372–381.
doi:10.1016/j.asoc.2011.08.035
Kaveh, A., & Javadi, M. S. (2014). Shape and size optimization of trusses with mul-
tiple frequency constraints using harmony search and ray optimizer for enhancing
the particle swarm optimization algorithm. Acta Mechanica, 225(6), 1595–1605.
doi:10.1007/s00707-013-1006-z
Khalili, M., Kharrat, R., Salahshoor, K., & Sefat, M. H. (2014). Global Dynamic
Harmony Search algorithm: GDHS. Applied Mathematics and Computation, 228,
195–219. doi:10.1016/j.amc.2013.11.058
Khazali, A. H., & Kalantar, M. (2011). Optimal reactive power dispatch based on
harmony search algorithm. International Journal of Electrical Power & Energy
Systems, 33(3), 684–692. doi:10.1016/j.ijepes.2010.11.018
Kong, X., Gao, L., Ouyang, H., & Li, S. (2015a). A simplified binary harmony
search algorithm for large scale 0–1 knapsack problems. Expert Systems with Ap-
plications, 42(12), 5337–5355. doi:10.1016/j.eswa.2015.02.015
Kong, X., Gao, L., Ouyang, H., & Li, S. (2015b). Solving large-scale multidimen-
sional knapsack problems with a new binary harmony search algorithm. Computers
& Operations Research, 63, 7–22. doi:10.1016/j.cor.2015.04.018

31
 Harmony Search Algorithm

Kougias, I. P., & Theodossiou, N. P. (2013). Multiobjective pump scheduling op-

timization using harmony search algorithm (HSA) and polyphonic HAS. Water
Resources Management, 27(5), 1249–1261. doi:10.1007/s11269-012-0236-5
Kulluk, S., Ozbakir, L., & Baykasoglu, A. (2012). Training neural networks with
harmony search algorithms for classification problems. Engineering Applications
of Artificial Intelligence, 25(1), 11–19. doi:10.1016/j.engappai.2011.07.006
Kumar, V., Chhabra, J. K., & Kumar, D. (2014). Parameter adaptive harmony search
algorithm for unimodal and multimodal optimization problems. Journal of Compu-
tational Science, 5(2), 144–155. doi:10.1016/j.jocs.2013.12.001
Layeb, A. (2013). A hybrid quantum inspired harmony search algorithm for 0–1
optimization problems. Journal of Computational and Applied Mathematics, 253,
14–25. doi:10.1016/j.cam.2013.04.004
Lee, S., & Mun, S. (2014). Improving a model for the dynamic modulus of asphalt
using the modified harmony search algorithm. Expert Systems with Applications,
41(8), 3856–3860. doi:10.1016/j.eswa.2013.12.021
Li, J., & Duan, H. (2014). Novel biological visual attention mechanism via Gauss-
ian harmony search. International Journal for Light and Electron Optics, 125(10),
2313–2319. doi:10.1016/j.ijleo.2013.10.075
Li, Y., Li, X., & Gupta, J. N. D. (2015). Solving the multi-objective flowline manu-
facturing cell scheduling problem by hybrid harmony search. Expert Systems with
Applications, 42(3), 1409–1417. doi:10.1016/j.eswa.2014.09.007
Luo, Q., Wu, J., Sun, X., Yang, Y., & Wu, J. (2012). Optimal design of groundwater
remediation systems using a multi-objective fast harmony search algorithm. Hydro-
geology Journal, 20(8), 1497–1510. doi:10.1007/s10040-012-0900-0
Luo, Q., Wu, J., Yang, Y., Qian, J., & Wu, J. (2014). Optimal design of groundwa-
ter remediation system using a probabilistic multi-objective fast harmony search
algorithm under uncertainty. Journal of Hydrology (Amsterdam), 519, 3305–3315.
doi:10.1016/j.jhydrol.2014.10.023
Mahdavi, M., & Abolhassani, H. (2009). Harmony K-means algorithm for document
clustering. Data Mining and Knowledge Discovery, 18(3), 370–391. doi:10.1007/
s10618-008-0123-0
Mahdavi, M., Fesanghary, M., & Damangir, E. (2007). An improved harmony search
algorithm for solving optimization problems. Applied Mathematics and Computa-
tion, 188(2), 1567–1579. doi:10.1016/j.amc.2006.11.033

32
Harmony Search Algorithm

Maheri, M. R., & Narimani, M. M. (2014). An enhanced harmony search algorithm

for optimum design of side sway steel frames. Computers & Structures, 136, 78–89.
doi:10.1016/j.compstruc.2014.02.001
Manahiloh, K. N., Nejad, M. M., & Momeni, M. S. (in press). Optimization of
design parameters and cost of geosynthetic-reinforced earth walls using harmony
search algorithm. Int. J. of Geosynth. and Ground Eng.
Manjarres, D., Del Ser, J., Gil-Lopez, S., Vecchio, M., Landa-Torres, I., Salcedo-Sanz,
S., & Lopez-Valcarce, R. (2013). On the design of a novel two-objective harmony
search approach for distance- and connectivity-based localization in wireless sen-
sor networks. Engineering Applications of Artificial Intelligence, 26(2), 669–676.
doi:10.1016/j.engappai.2012.06.002
Merzougui, A., Hasseine, A., & Laiadi, D. (2012). Application of the harmony search
algorithm to calculate the interaction parameters in liquid–liquid phase equilibrium
modeling. Fluid Phase Equilibria, 324, 94–101. doi:10.1016/j.fluid.2012.03.029
Miguel, L. F. F., Miguel, L. F. F., Kaminski, J. Jr, & Riera, J. D. (2012). Damage
detection under ambient vibration by harmony search algorithm. Expert Systems
with Applications, 39(10), 9704–9714. doi:10.1016/j.eswa.2012.02.147
Morsali, R., Jafari, T., Ghods, A., & Karimi, M. (2014). Solving unit commitment
problem using a novel version of harmony search algorithm. Front. Energy, 8(3),
297–304.
Mukhopadhyay, A., Roy, A., Das, S., & Abraham, A. (2008). Population-variance
and explorative power of harmony search: An analysis.Second National Conference
on Mathematical Techniques Emerging Paradigms for Electronics and IT Industries
(MATEIT 2008). doi:10.1109/ICDIM.2008.4746793
Mun, S., & Cho, Y.-H. (2012). Modified harmony search optimization for constrained
design problems. Expert Systems with Applications, 39(1), 419–423. doi:10.1016/j.
eswa.2011.07.031
Murren, P., & Khandelwal, K. (2014). Design-driven harmony search (DDHS)
in steel frame optimization. Engineering Structures, 59, 798–808. doi:10.1016/j.
engstruct.2013.12.003
Nekkaa, M., & Boughaci, D. (2016). Hybrid harmony search combined with sto-
chastic local search for feature selection. Neural Processing Letters, 44(1), 199–220.
doi:10.1007/s11063-015-9450-5

33
 Harmony Search Algorithm

Niu, Q., Zhang, H., Wang, X., Li, K., & Irwin, G. W. (2014). A hybrid harmony search
with arithmetic crossover operation for economic dispatch. International Journal of
Electrical Power & Energy Systems, 62, 237–257. doi:10.1016/j.ijepes.2014.04.031
Omran, M. G. H., & Mahdavi, M. (2008). Global-best harmony search. Applied
Mathematics and Computation, 198(2), 643–656. doi:10.1016/j.amc.2007.09.004
Pandi, V. R., & Panigrahi, B. K. (2011). Dynamic economic load dispatch using
hybrid swarm intelligence based harmony search algorithm. Expert Systems with
Applications, 38(7), 8509–8514. doi:10.1016/j.eswa.2011.01.050
Papa, J. P., Scheirer, W., & Cox, D. D. (2016). Fine-tuning Deep Belief Networks
using Harmony Search. Applied Soft Computing, 46, 875–885. doi:10.1016/j.
asoc.2015.08.043
Paqaleh, M. A., Rashidinejad, M., & Kasmaei, M. P. (2010). An implementation
of harmony search algorithm to unit commitment problem. Electrical Engineering,
92(6), 215–225. doi:10.1007/s00202-010-0177-z
Pires, R. G., Pereira, D. R., Pereira, L. A. M., Mansano, A. F., & Papa, J. P. (2016).
Projections onto convex sets parameter estimation through harmony search and its
application for image restoration. Natural Computing, 15(3), 493–502. doi:10.1007/
s11047-015-9507-4
Purnomo, H. D., & Wee, H.-M. (2014). Maximizing production rate and workload
balancing in a two-sided assembly line using Harmony Search. Computers & In-
dustrial Engineering, 76, 222–230. doi:10.1016/j.cie.2014.07.010
Rastgou, A., & Moshtagh, J. (2014). Improved harmony search algorithm for trans-
mission expansion planning with adequacy–security considerations in the deregulated
power system. International Journal of Electrical Power & Energy Systems, 60,
153–164. doi:10.1016/j.ijepes.2014.02.036
Razfar, M. R., Zinati, R. F., & Haghshenas, M. (2011). Optimum surface roughness
prediction in face milling by using neural network and harmony search algorithm.
International Journal of Advanced Manufacturing Technology, 52(5-8), 487–495.
doi:10.1007/s00170-010-2757-5
Sakthivel, V., Elias, E. (2015). Design of low complexity sharp MDFT filter banks
with perfect reconstruction using hybrid harmony-gravitational search algorithm.
Engineering Science and Technology, an International Journal, 18(4), 648-657..

34
Harmony Search Algorithm

Salcedo-Sanz, S., Manjarrés, D., Pastor-Sánchez, Á., Del Ser, J., Portilla-Figueras,
J. A., & Gil-López, S. (2013). One-way urban traffic reconfiguration using a multi-
objective harmony search approach. Expert Systems with Applications, 40(9),
3341–3350. doi:10.1016/j.eswa.2012.12.043
Samiee, M., Amjady, N., & Sharifzadeh, H. (2013). Security constrained unit com-
mitment of power systems by a new combinatorial solution strategy composed of
enhanced harmony search algorithm and numerical optimization. International
Journal of Electrical Power & Energy Systems, 44(1), 471–481. doi:10.1016/j.
ijepes.2012.07.069
Segura, T. G., Yepes, V., Alcalá, J., & Pérez-López, E. (2015). Hybrid harmony
search for sustainable design of post-tensioned concrete box-girder pedestrian
bridges. Engineering Structures, 92, 112–122. doi:10.1016/j.engstruct.2015.03.015
Shivaie, M., Ameli, M. T., Sepasian, M. S., Weinsier, P. D., & Vahidinasab, V.
(2015). A multistage framework for reliability-based distribution expansion planning
considering distributed generations by a self-adaptive global-based harmony search
algorithm. Reliability Engineering & System Safety, 139, 68–81. doi:10.1016/j.
ress.2015.03.001
Sinsuphan, N., Leeton, U., & Kulworawanichpong, T. (2013). Optimal power flow
solution using improved harmony search method. Applied Soft Computing, 13(5),
2364–2374. doi:10.1016/j.asoc.2013.01.024
Sirjani, R., & Bade, M. G. (2015). A global harmony search algorithm for finding
optimal capacitor location and size in distribution networks. J. Cent. South Univ,
22(5), 1748–1761. doi:10.1007/s11771-015-2693-5
Sivasubramani, S., & Swarup, K. S. (2011). Multi-objective harmony search algo-
rithm for optimal power flow problem. International Journal of Electrical Power
& Energy Systems, 33(3), 745–752. doi:10.1016/j.ijepes.2010.12.031
Taleizadeh, A. A., Niaki, S. T. A., & Seyedjavadi, S. M. H. (2012). Multi-product
multi-chance-constraint stochastic inventory control problem with dynamic demand
and partial back-ordering: A harmony search algorithm. Journal of Manufacturing
Systems, 31(2), 204–213. doi:10.1016/j.jmsy.2011.05.006
Tuo, S., Zhang, J., Yong, L., Yuan, X., Liu, B., Xu, X., & Deng, F. A. (2015). A
harmony search algorithm for high-dimensional multimodal optimization problems.
Digital Signal Processing, 46, 151–163. doi:10.1016/j.dsp.2015.08.008

35
 Harmony Search Algorithm

Valian, E., Tavakoli, S., & Mohanna, S. (2014). An intelligent global harmony
search approach to continuous optimization problems. Applied Mathematics and
Computation, 232, 670–684. doi:10.1016/j.amc.2014.01.086
Wang, C. M., & Huang, Y. F. (2010). Self-adaptive harmony search algorithm for
optimization. Expert Systems with Applications, 37(4), 2826–2837. doi:10.1016/j.
eswa.2009.09.008
Wang, G.-G., Gandomi, A. H., Zhao, X., & Chu, H. C. E. (2015a). Hybridizing
harmony search algorithm with cuckoo search for global numerical optimization.
Soft Computing, 20(1), 273–285. doi:10.1007/s00500-014-1502-7
Wang, L., & Li, L. P. (2013). An effective differential harmony search algorithm for
the solving non-convex economic load dispatch problems. International Journal of
Electrical Power & Energy Systems, 44(1), 832–843. doi:10.1016/j.ijepes.2012.08.021
Wang, Y., Liu, Y., Feng, L., & Zhu, X. (2015b). Novel feature selection method
based on harmony search for email classification. Knowledge-Based Systems, 73,
311–323. doi:10.1016/j.knosys.2014.10.013
Zheng, L., Diao, R., & Shen, Q. (2015a). Self-adjusting harmony search-based fea-
ture selection. Soft Computing, 19(6), 1567–1579. doi:10.1007/s00500-014-1307-8
Zheng, Y.-J., Zhang, M.-X., & Zhang, B. (2015b). Biogeographic harmony search
for emergency air transportation. Soft Computing, 20(3), 967–977. doi:10.1007/
s00500-014-1556-6
Zinati, R. F., & Razfar, M. R. (2012). Constrained optimum surface roughness pre-
diction in turning of X20Cr13 by coupling novel modified harmony search-based
neural network and modified harmony search algorithm. International Journal of
Advanced Manufacturing Technology, 58(1-4), 93–107. doi:10.1007/s00170-011-
3393-4

36

View publication stats