A Comprehensive Review of the Fireworks Algorithm

JUNZHI LI and YING TAN, Peking University, China

The fireworks algorithm, which is inspired from the phenomenon of fireworks explosion, is a special kind of
swarm intelligence algorithm proposed in 2010. Since then, it has been attracting more and more research
interest and has been widely employed in many real-world problems due to its unique search manner and high
efficiency. In this article, we present a comprehensive review of its advances and applications. We begin with
an introduction to the original fireworks algorithm. Then we review its algorithmic research work for single
objective and multi-objective optimization problems. After that, we present the theoretical analyses of the
fireworks algorithm. Finally, we give a brief overview of its applications and implementations. Hopefully, this
article could provide a useful road map for researchers and practitioners who are interested in this algorithm
and inspire new ideas for its further development.
CCS Concepts: • Theory of computation → Evolutionary algorithms; Bio-inspired optimization;
Additional Key Words and Phrases: Fireworks algorithm, evolutionary computation, swarm intelligence
ACM Reference format:
Junzhi Li and Ying Tan. 2019. A Comprehensive Review of the Fireworks Algorithm. ACM Comput. Surv. 52,
6, Article 121 (November 2019), 28 pages.
https://doi.org/10.1145/3362788

1 INTRODUCTION
Optimization is one of the most fundamental and common problems in science and industry. Con-
vex optimization with available derivatives can be solved efficiently by gradient-based algorithms
like gradient descent or the Newton’s method. However, there is no derivative information in
black-box or discrete optimization problems, and gradient-based algorithms are often trapped in
local optima when optimizing multi-modal objective functions even though there is derivative in-
formation. Therefore, a new branch of approaches, i.e., evolutionary computation was proposed
and developed, which do not require derivative information and can keep a balance between ex-
ploitation and exploration in the global search process. Nowadays, evolutionary computation has
become one of the most active subfield of artificial intelligence.
Evolutionary computation in a narrow sense refer to algorithms that are directly inspired by bi-
ological evolutionary phenomena, such as evolutionary strategy and the genetic algorithm. While
evolutionary computation in a broad sense refer to all derivative-free iterative optimization al- 121
gorithms, including metaphor-based algorithms (meta-heuristics) and direct methods such as the
downhill simplex method and estimation of distribution algorithms. So far, more than one hundred

This work was supported by the Natural Science Foundation of China (NSFC) under grant no. 61673025 and partially
supported by National Key Basic Research Development Plan (973 Plan) Project of China under grant no. 2015CB352302.
Authors’ address: J. Li and Y. Tan, Peking University, Beijing, China; emails: {ljz, ytan}@pku.edu.cn.
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee
provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and
the full citation on the first page. Copyrights for components of this work owned by others than ACM must be honored.
Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires
prior specific permission and/or a fee. Request permissions from permissions@acm.org.
© 2019 Association for Computing Machinery.
0360-0300/2019/11-ART121 $15.00
https://doi.org/10.1145/3362788

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
121:2 J. Li and Y. Tan

different evolutionary algorithms have been proposed to solve different kinds of optimization prob-
lems [200].
Swarm intelligence algorithms (e.g., particle swarm optimization (PSO) [81], artificial bee colony
(ABC) [78], and ant colony optimization (ACO) [44]) are a special group of evolutionary algorithms
that are featured by various effective cooperative mechanisms among individuals within the pop-
ulation. However, evolutionary algorithms in a narrow sense (e.g., genetic algorithm (GA) [55] and
evolution strategy (ES) [127]) accomplish global optimization by competitive mechanisms inspired
by the natural selection process. To combine the advantages of both branches, the fireworks algo-
rithm has been proposed. The fireworks algorithm employs a unique framework allowing both co-
operative and competitive mechanisms and a novel search manner called explosion. It has been at-
tracting more and more research interest due to its unique design and high efficiency. In this article,
we present a comprehensive review of the fireworks algorithm for its researchers and practitioners.

2 FIREWORKS ALGORITHM
The fireworks algorithm (FWA) is a new kind of global optimization algorithm proposed by Tan
and Zhu [172] in 2010. It conducts the global search by mimicking the phenomenon of fireworks
explosion. The algorithm is featured by (1) an explosive search manner and (2) a framework en-
abling multiple (sub)populations to interact. Since it was proposed, the FWA has attracted much
research interest and has been widely applied in real-world optimization problems. In this article,
the fireworks algorithm proposed in 2010 is referred to as “the original fireworks algorithm.”
Without loss of generality, the following continuous minimization problem is considered in this
section:
minx∈Rd f (x), (1)
where x represents a vector in the d-dimensional Euclidean space. The purpose of the algorithm
is to locate the optimal x∗ , which means it is of the minimal evaluation/fitness value f (x∗ ).

2.1 Framework
The original FWA repeats the following steps until the terminal criterion (required precision, max-
imal function evaluations, etc.) is met. (1) At the beginning, randomly choose some locations in
the search space as the initial locations of the fireworks. (2) Conduct explosive search around the
locations of the fireworks to generate explosion sparks. (3) Mutate the fireworks by certain rules
to generate mutation sparks. (4) Select the fireworks of the next generation from the current fire-
works and sparks.
Many improvements have been made upon the operators in the FWA, yet most variants gener-
ally follow the framework introduced above. Therefore, this can be considered as the basic frame-
work of the FWA.
There are four main operators in the original fireworks algorithm: the explosion operator, the
mutation operator, the selection operator and the mapping rule.

2.2 Explosion
The fireworks algorithm is inspired by the phenomenon of fireworks explosion. Although there
are essential differences between the algorithm and the real fireworks, the metaphor captures
the core feature of the explosive search manner of the algorithm. The explosion operator is the
basic search manner in the FWA. In the explosion operator, a certain number of explosion sparks
are generated within a certain explosion amplitude around each firework. In the original FWA, if
the fitness of a firework is relatively good, then more explosion sparks will be generated within
a smaller explosion amplitude around it to conduct local exploitation. While, if the fitness of a

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
A Comprehensive Review of the Fireworks Algorithm 121:3

firework is relatively bad, then fewer explosion sparks will be generated within a larger explosion
amplitude around it to conduct global exploration. The fireworks interchange the information of
fitness values to determine the allocation of explosion sparks’ numbers and explosion amplitudes.
In the original FWA, the number of explosion sparks of each firework is determined by the
following equation:
maxk ( f (Xk )) − f (Xi )
λi = λ ·  , (2)
j (maxk ( f (Xk )) − f (X j ))

where Xi is the location of the ith firework. Obviously, λ = i λi . Therefore, λ is a parameter to
control the total number of explosion sparks in each generation. The smaller f (Xi ) is, the more
explosion sparks are generated. To avoid the overwhelming effect of good fireworks, thresholds
are set in the original FWA:
⎧
⎪
⎪ round(aλ) if λi < aλ
⎪
λi = ⎨⎪ round(bλ) if λi > bλ , (3)
⎪
⎪ round(λ ) otherwise
⎩ i

where round(x ) means the integer closest to x, a and b (a < b < 1) are two parameters to control
the lower and upper boundaries of explosion sparks’ number.
The explosion amplitude of each firework is determined by the following equation:
f (Xi ) − mink ( f (Xk ))
Ai = A ·  . (4)
j ( f (X j ) − mink ( f (Xk )))

Likewise, A is a parameter to control the total explosion amplitude. The smaller f (Xi ), the smaller
the amplitude.
Algorithm 1 shows how the explosion sparks of the ith firework are generated, where d is the
dimensionality of the search space and si jk is the kth dimension of the jth explosion spark of the
ith firework.

2.3 Mutation
Aside from explosion sparks, a number of mutation sparks are also generated in the original FWA
by the mutation operator, as shown in Algorithm 2, where λ̇ is the number of mutation sparks.

2.4 Mapping Rule

Real-world optimization problems and standard benchmark suites are often with boundary
constraints x ∈ [lb, ub], which is a simple kind of constraint compared with other equality or
inequality constraints. In the FWA, newly generated explosion or mutation sparks may be located
outside the boundaries, in which case they need to be replaced by ones within the boundaries,
otherwise function evaluations may be wasted.

ALGORITHM 1: Explosion Operator

for j = 1 to λi do
e ← Ai · rand(−1, 1)
randomly choose z dimensions, z ← round(d · rand(0, 1))
si j ← Xi
for each chosen dimension k do
si jk ← si jk + e
end
end

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
121:4 J. Li and Y. Tan

ALGORITHM 2: Mutation Operator

for j = 1 to λ̇ do
randomly choose a firework Xi
e ← randn(1, 1)
randomly choose z dimensions, z ← round(d · rand(0, 1))
Gj ← Xi
for each chosen dimension k do
G jk ← G jk · e
end
end

In the original FWA, a modular arithmetic-based mapping rule is used. If an individual x satisfies
x k < lbk or x k > ubk in dimension k, then
x k ← lbk + |x k | mod (ubk − lbk ). (5)

2.5 Selection
In every generation, after the sparks are generated, it is necessary to select the fireworks of the
next generation. In the original FWA, a density-based selection mechanism is proposed, intending
to maintain the representativeness and the diversity of selected fireworks.
First, the candidate (including current fireworks and sparks) of the best fitness value is selected
as a firework.
Second, the rest μ − 1 fireworks are selected from the rest candidates. The probability each can-
didate x i is selected is
R(xi )
p(xi ) =  , (6)
x j ∈K R(x j )
 
where R(xi ) = xj ∈K d (xi , xj ) = xj ∈K ||xi − xj ||, and K represents the set consisting of all cur-
rent fireworks and sparks.
The fireworks algorithm is originally considered as a swarm intelligence algorithm. However, its
framework is quite different from other typical swarm intelligence algorithms, in which the popu-
lation size is usually fixed (like PSO or ACO) or at least seems so (like ABC). While in evolutionary
algorithms, the numbers of parents and offspring are different (like GA or ES). From this point of
view, the FWA is more similar to evolutionary algorithms. However, it is also different from evo-
lutionary algorithms, because it contains the idea that multiple fireworks interact and cooperate
to accomplish global optimization. Therefore, the FWA should be considered as a developmental
(evolutionary) swarm intelligence algorithm [157].
Tan et al. published an introduction [171] and a monograph [170] of the FWA in 2013 and 2015,
respectively, which introduced the principles and typical applications of the algorithm in detail.
While in this article, we summarize the literature of the FWA from 2010 to 2018. This is the first
comprehensive review of the fireworks algorithm.

3 ALGORITHMIC RESEARCH
In this section, we review the literature about the algorithmic research of the FWA, including
improvements of the FWA itself, hybridizations of the FWA and other algorithms, and adaptations
of the FWA for multi-objective optimization or dynamic optimization problems. However, FWA
variants for solving discrete optimization problems are reviewed along with the applications in
Section 5, because in discrete/combinatorial optimization problems, how solutions are encoded is
more problem related.

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
A Comprehensive Review of the Fireworks Algorithm 121:5

Fig. 1. Dendrogram of FWA variants.

3.1 Algorithmic Improvements

Figure 1 shows an overview of the variants of the FWA.
3.1.1 Operator Improvements. Zheng et al. [237] conducted a quite thorough analyses of the
original FWA.
(1) According to the explosion operator in the original FWA (Algorithm 1), the displacement
e remains unchanged in every dimension for a certain spark, which may limit the explo-
ration ability of the algorithm. Therefore, they proposed that the displacement should take
different random numbers in each dimension.
(2) According to Equation (4), the explosion amplitude of the firework with the best fitness is
0, which makes it unable to search effectively. Therefore, they proposed a minimal explo-
sion amplitude check strategy to prevent the explosion amplitudes from being too small, in
which the threshold of the explosion amplitudes is a (linearly or non-linearly) decreasing
function of the number of evaluations.
(3) The mutation operator and the mapping rule in the original FWA are concentrated near
the origin. Therefore, the performance may suffer greatly when the optimal point of the
objective function is moved away from the origin. Therefore, they proposed a new kind
of mutation operator and suggested a random mapping rule.
(4) The density-based selection operator is time-consuming. So they proposed replacing it
with an elitism-random selection operator.
Based on these analyses and improvements, the enhanced fireworks algorithm (EFWA) is pro-
posed. Experimental results on 12 different test functions with different shift values indicate that
the performance of the EFWA is stable even when the objective function is shifted and that the
computational cost of the EFWA is less than the original FWA.
Liu et al. [111] pointed out by experiments that the explosion amplitudes and the explosion
sparks’ number are unstable in the original FWA. They proposed using a rank-based transfer func-
tion to calculate Ai and λi . Besides, they, too, realized the problem in the mutation operator of the

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
121:6 J. Li and Y. Tan

original FWA and therefore suggested using a pure-random mutation operator. They also pointed
out that the density-based selection operator only considers the distance but not the fitness val-
ues of individuals, and therefore they proposed two kinds of alternatives: using a fitness-based
roulette or greedily selecting the best ones. Experimental results on 14 of the functions of the
CEC2005 benchmark indicate that the greedy selection outperforms the roulette and that the pro-
posed improved FWA (IFWA) outperforms the original FWA, and its performance is comparable
to PSO. Si et al. [159] proposed a new kind of transfer function based on the IFWA, in which not
only the ranks but also the differences of fireworks’ fitness are used. Experimental results on the
CEC2013 benchmark indicate that the proposed algorithm (FWA-ATF) outperforms the original
FWA, the EFWA, and the IFWA significantly.
Li et al. [100] and Zheng et al. [236] respectively proposed two different approaches to control
the explosion amplitude adaptively (dynamically). They pointed out that the threshold in the min-
imal explosion amplitude check strategy in the EFWA is a manually set function, which cannot
adapt to different objective functions and different search phases. Therefore, they proposed adap-
tive FWA (AFWA) and dynamic search FWA (dynFWA), respectively. Their core ideas and effects
are similar [238]: In each generation, if a better solution is found, then the amplitude should be
amplified; otherwise, it should be reduced. Besides, Zheng et al. also pointed out that the mutation
operator in the EFWA is not effective in the dynFWA, and thus removed it. Experimental results
on the 28 functions of the CEC2013 benchmark [108] indicate that both algorithms significantly
outperform the EFWA and several PSO variants.
Zhang et al. [225] proposed a new mutation operator and a new selection operator based on
the EFWA. In the new mutation operator, two individuals are chosen randomly, and a mutation
spark is generated on the line connecting them. In the new selection operator, for each candidate,
q opponents are randomly chosen from the whole population, and the fireworks of the next gen-
eration are the candidates that win most times. Experimental results on 18 test functions indicate
that both improvements are effective and the proposed improved EFWA (IEFWA) outperforms the
EFWA significantly. Later, they [227] hybridized the migration operator of bio-geography-based
optimization (BBO) [160] with the explosion operator in the FWA. Experimental results on the
CEC2015 benchmark [107] indicate that the EFWA with the three strategies (FWA_EI) outper-
forms the EFWA and BBO significantly.
Cheng et al. [36] too pointed out the problem of the mapping rule in the original FWA. There-
fore, they tested four different kinds of mapping rules. Experimental results on shifted objective
functions indicate that mapping to the boundary and mapping to limited stochastic region out-
perform other mapping rules. Ye et al. [212] explored the same topic in large-scale optimization.
They found that the mirror mapping rule outperforms the others on most test functions.
Li et al. [104] proposed five improvements in the FWA. (1) In the initialization step, the fireworks
and their opposition solutions are together evaluated, the better ones are taken as the fireworks in
the first generation. (2) A dynamic explosion amplitude strategy is proposed, which is the same as
in the dynFWA except that when all the fireworks are of the same fitness, the explosion amplitude
will also be amplified. (3) The Gaussian distribution used in the mutation operator in the EFWA
is replaced with a t-distribution. (4) The opposition solution of the best firework is introduced in
the population as an elite solution. (5) Other than the best candidate, other fireworks are selected
based on a special kind of roulette, in which good and bad candidates are of higher probability but
medium candidates are of lower probability to be selected. The proposed algorithm proves to be
globally convergent. Experimental results on the CEC2013 benchmark indicate that the proposed
algorithm outperforms the original FWA, the EFWA, and the dynFWA significantly.
Li et al. [103] proposed a new displacement method based on the AFWA (which is unfortunately
based on a misunderstanding of the AFWA) and an elitism-tournament selection operation. In the

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
A Comprehensive Review of the Fireworks Algorithm 121:7

elitism-tournament selection operation, the best candidate is first selected. Then, each time two
individuals are randomly chosen from the rest candidates, the better one is selected a firework.
This process is repeated for μ − 1 times. Experimental results on the CEC2015 benchmark indicate
that the proposed TMSFWA outperforms the EFWA, the AFWA, and the dynFWA.
Li et al. [105] introduced an adaptive mutation into the dynFWA, i.e., using Lévy distribution in
early phases and using Gaussian distribution in late phases. Experimental results on the CEC2013
benchmark indicate that the proposed algorithm (AMdynFWA) outperforms the EFWA, the AFWA,
and the dynFWA significantly.
Yu and Takagi [220] proposed two improvements upon the original FWA. (1) The amplitudes
are no longer allocated according to fitness values. Instead, the amplitudes of all fireworks are set
to the same value that linearly decreases until a certain time, and remains unchanged after that.
(2) A new selection mechanism similar to the independent selection is proposed, but the mutation
operator is adopted and the mutation sparks are combined with the candidate pool of the worst
firework. Experimental results on 20 of the test functions of the CEC2013 benchmark indicate that
both proposed strategies are effective and the proposed algorithm outperforms the original FWA
significantly.
Li and Tan [98] proposed a simplified version of the dynFWA, called bare bones FWA (BBFWA).
In the BBFWA, only one firework is adopted, and the mutation operator and the dimension selec-
tion mechanism are all removed. Experimental results on the CEC2013 benchmark indicate that
the BBFWA outperforms the EFWA, the AFWA, the dynFWA, the CoFFWA, and several other
evolutionary algorithms, and the complexity of the BBFWA is smaller than these algorithms. Its
performance on three of the real-world problems of the CEC2011 competition [38] is comparable
to the champion of that competition.
Yu et al. [223] proposed generating sparks one by one, and each spark is generated around the
former one except that the first one is generated around the firework. However, once the fitness
of the new spark is worse than the former, this process stops and starts over from the firework.
They also proposed a new selection strategy, which is different from the one in the EFWA in
that the sparks whose fitness are worse than the firework are no longer considered as candidates.
Experimental results on the CEC2013 benchmark indicate the proposed strategies improve the
performance of the EFWA significantly.
Cheng et al. [35] proposed three improvements based on the EFWA. (1) They proposed a new
explosion operator that can allegedly generate sparks in a spherical neighborhood of each firework.
(2) They proposed a new mutation operator that allows all fireworks and sparks to have equal
chances to generate mutation sparks. (3) A deep information exchange strategy that is borrowed
from the Grey Wolf Optimizer (GWO) [128] is introduced to act on the candidate set. Experimental
results on 23 test functions indicate the proposed algorithm outperforms the EFWA and several
other opponents.
Guo and Liu [65] replaced the Gaussian sparks in the EFWA with a new kind of sparks that are
generated around the middle point between the best fireworks and a randomly selected firework.
Experimental results on 20 thirty-dimensional test functions indicate that the proposed improved
EFWA outperforms both the EFWA and the dynFWA significantly.
Yu et al. [222] proposed a novel explosion paradigm called multi-layer explosion. In each explo-
sion operation, after sparks are generated and evaluated, each spark will generate a certain number
of subsequent sparks within certain amplitudes depending on the fitness of these sparks. In this
way, more landscape information is utilized and the shape of the explosion becomes more flexible.
Experimental results on the CEC2013 test suite indicate that the performance of the EFWA can be
significantly improved using the proposed paradigm.

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
121:8 J. Li and Y. Tan

3.1.2 Elite Strategies. Pei et al. [138] introduced an elite strategy into the FWA. In every gener-
ation, a number of individuals are chosen to conduct fitness landscape approximation on each pro-
jected dimension. Then a promising elite spark is generated using the approximation. If its fitness
is better than the worst individual, then the worst individual will be replaced with the elite spark.
Experimental results on 10 of the test functions of the CEC2005 benchmark suite [166] indicate
the effectiveness of the proposed method. Later, they [139] investigated different kinds of approxi-
mation approaches. Experimental results on all test functions of the CEC2005 benchmark indicate
that approximation in low dimensional space is better than in higher dimensional or the original
space, and individuals used for approximation should be chosen randomly from the population.
Li et al. [96, 101] pointed out that the information obtained by the explosion sparks is not fully
used in the dynFWA. Therefore, they proposed an information utilization-based guiding mutation
operator. By calculating the difference vector between good explosion sparks and bad explosion
sparks, which is used to lead the movement of the fireworks, the convergence speed and the ex-
ploration capability can be enhanced. Experimental results on the CEC2013 benchmark indicate
that the dimension selection mechanism in the explosion operator is not useful in the proposed
guided FWA (GFWA), and the GFWA outperforms the EFWA, the AFWA, the dynFWA, and sev-
eral other evolutionary algorithms significantly. Experimental results on the CEC2010 large-scale
optimization benchmark [173] indicate that the performance of GFWA on large-scale problems
(1,000 dimensions) is comparable to state-of-the-art large-scale optimization algorithms.
Yu et al. [221] proposed another elite strategy based on the EFWA. First, each firework and
the sparks it generated are used to calculate a gradient-like vector, which is a weighted sum of
the differences between the firework and each spark. Second, the vectors are used to estimate a
convergence point, which will be finally introduced into the population and take the place of the
worst candidate if its fitness is better. Experimental results on 20 of the CEC2013 test functions
indicate the performance of the proposed algorithm outperforms the EFWA significantly.

3.1.3 Interactive Mechanisms. Zheng et al. [238] showed experimentally that the fireworks
other than the best one contribute little to the search process, because the fireworks are se-
lected from the same candidate pool, which makes the information of bad fireworks not inherited.
Therefore, they proposed an independent selection framework, in which each firework is selected
from its own offsprings. Besides, they also proposed a crowdedness-avoiding strategy to prevent
other fireworks to search the same local area with the best firework. Experimental results on the
CEC2013 benchmark indicate that the mutation operation in the EFWA is not useful either in the
dynFWA or the AFWA, and thus it is removed in the proposed cooperative framework FWA (CoF-
FWA). Judging from the average ranks, the CoFFWA outperforms the EFWA, the AFWA, and the
dynFWA and several other swarm intelligence algorithms.
Zhao et al. [235] proposed three improvements based on the dynFWA. (1) The explosion sparks’
number of the best firework is no longer calculated according to Equation (2). Instead, it is set to
a fixed number. (2) The other fireworks will be reinitialized with a certain probability and move
toward the best one. (3) A mirror mapping rule is adopted instead of the random mapping rule.
Experimental results on the CEC2015 benchmark indicate that the proposed elite-leading FWA
(ELFWA) outperforms the dynFWA, the dynFWACM, and the eddynFWA.
Laña et al. [86] introduced a wind inertia dynamics into the EFWA to force generated sparks
to move toward the best firework. Experimental results on six test functions indicate that the
proposed algorithm (EFWA-WID) outperforms the EFWA significantly.
Li and Tan [97, 99] proposed a loser-out tournament-based fireworks algorithm based on the
independent selection framework, in which if the fitness of a firework cannot catch up with the

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
A Comprehensive Review of the Fireworks Algorithm 121:9

best one, it will reinitialized to elevate the probability of avoiding local optima. Experimental re-
sults on the CEC2013 benchmark indicate that the loser-out tournament-based FWA (LoTFWA)
outperforms the EFWA, the AFWA, the dynFWA, the CoFFWA, and the GFWA on multi-modal
test functions significantly, and, according to average ranks, the LoTFWA also outperforms restart
CMA-ES with increasing population (IPOP-CMAES) [9] on multi-modal test functions.

3.1.4 Other Work. Zheng et al. [241] showed by experiments based on the dynFWA that with
smaller z (i.e., fewer explosion dimensions) in the explosion operator, the generated explosion
sparks have a greater chance to surpass the firework. Therefore, they proposed reducing the num-
ber of explosion dimensions by a certain coefficient after every certain number of generations. Ex-
perimental results of the proposed algorithm (ed-dynFWA) on the CEC2015 benchmark are shown,
but no comparison was made against other algorithms.
Chen et al. [30] proposed an approach to balance exploration and exploitation in the original
FWA by utilizing landscape information. If the coverage of the sparks is large, then the sparks are
generated randomly in the next generation. If the coverage is middle, then sparks are generated by
utilizing landscape information. Otherwise, sparks are generated in the original way. Experimental
results on eight test functions indicate the proposed algorithm outperforms the original FWA and
the EFWA.
Kumar et al. [85] evaluated the original FWA on six test functions and obtained the optimal
parameter setting on these problems.
Barraza et al. [15, 16, 18] proposed using fuzzy logic to dynamically adjust the number of explo-
sion sparks and explosion amplitudes. According to different search phases, the total amplitude
and number of explosion sparks are subject to different membership functions. Experimental re-
sults on 12 test functions indicate that the proposed algorithm (IFFWA) outperforms the original
FWA significantly. Later, they [17] introduced a new kind of input into the fuzzy system called
sparks dispersion measure. Experimental results on 14 test functions with different shift values
indicate that the new algorithm (DPIFFWA) outperforms the IFFWA significantly.
Gong [56] proposed using chaotic maps to control the amplification coefficient in the AFWA.
Experimental results on 12 shifted test functions indicate that the circle map–based chaotic AFWA
(CAFWA) outperforms the AFWA significantly, and according to average ranks, the CAFWA out-
performs the original FWA, the EFWA, and several other swarm intelligence algorithms.
Zhang et al. [229] proposed a resampling-based FWA for noisy optimization problems. The core
idea is enlarging the number of top sparks resampled while reducing the number of resampling
times with the search process. Experimental results on the CEC2015 benchmark with different
noise levels indicate that the proposed algorithm (FWA-NO) outperforms CoFFWA and its simple
resampling versions.

3.2 Hybrid Algorithms

Yu et al. [215, 218] replaced the mutation operator in the EFWA with the mutation strategy in
differential evolution (DE). Experimental results on the CEC2014 benchmark [106] indicate that
the proposed EFWA with differential mutation (FWA-DM) outperforms the EFWA significantly.
Zheng et al. [243] proposed another kind of hybrid algorithm of the FWA and DE. After sparks
are generated, μ individuals are selected from the best individuals using roulette. Then DE opera-
tors are conducted upon the μ individuals to generate trail solutions, which will replace the original
ones if they are of better fitness. After the new μ solutions are the fireworks of the next generation.
Experimental results on eight test functions show the advantage of the hybrid algorithm over the
both components.

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
121:10 J. Li and Y. Tan

Yu et al. [216, 217, 219] successively introduced a covariance mutation into the AFWA, the dyn-
FWA, and the CoFFWA. Experimental results on the CEC2015 benchmark indicate that the AFWA
with covariance mutation (FWA-CM) outperforms the AFWA significantly.
Gao et al. [53] combines the FWA with the opposition-based learning and quantum computing
operators. Experimental results on five test functions indicate that the proposed opposition-based
quantum FWA (OQFWA) outperforms the original FWA.
Bacanin et al. [12] used the search manner in the firefly algorithm to replace the mutation op-
erator in the original fireworks algorithm. Experimental results on six test functions indicate that
the proposed algorithm outperforms the original FWA and several PSO variants, while Wang et al.
[192] introduce the explosion operator into the firefly algorithm to improve its local search. Ex-
perimental results on the CEC2013 suite indicate that the proposed hybrid algorithm outperforms
the FWA and variants of the firefly algorithm.
Gong [57] combines the AFWA with opposition-based learning. In the initialization step, the
fireworks and their opposition solutions are together evaluated, and the better ones are taken
as the fireworks in the first generation. In each generation, with a certain probability, the quasi
opposites of selected fireworks are also evaluated, and the fittest ones among fireworks and their
quasi opposites are the fireworks of the next generation. Experimental results on 12 test functions
indicate that the proposed opposition-based adaptive fireworks algorithm (OAFWA) outperforms
the AFWA significantly.
Sun et al. [168] introduced the grouping strategy of the shuffled frog leaping algorithm (SFLA)
[48] into the original FWA. Experimental results on four test functions indicate that the hybrid
algorithm outperforms both the original FWA and the SFLA.
Ye and Wen [211] proposed using the simulated annealing factor to replace the minimal explo-
sion amplitude check strategy proposed in EFWA. Experimental results on five of the test functions
of the CEC2013 benchmark and three of the test functions of the CEC2014 benchmark indicate that
the hybrid algorithm outperforms the double elite co-evolutionary genetic algorithm and the dif-
ferential evolution algorithm based on self-adapting mountain-climbing operator.
Chen [31] proposed a hybrid algorithm of PSO and the FWA, in which the operators of PSO
are used for exploration, while the operators of the FWA are used for exploitation. Experimental
results on 22 test functions indicate the hybrid algorithm (PS-FWA) outperforms PSO and the
original FWA.
Barraza et al. [19] proposed a hybrid algorithm of the FWA and GWO, in which the initialization
is conducted using the explosive manner. Experimental results on 22 test functions indicate the
hybrid algorithm performs well when the dimensionality is low.
Two different groups proposed two versions of hybridization of BBO and the FWA. Zhang et al.
[226] selected the operators of the EFWA and BBO by a certain probability in each iteration, while
Farswan and Bansal [49] used the operators of BBO and the original FWA in turn in each iteration.
Both hybridizations are evaluated and shown effective on a number of test functions.

3.3 Dynamic Optimization

Pekdemir and Topcuoglu [140] proposed two variants of the EFWA for dynamic optimization
problems. In the EFWA-D1, explosion sparks’ numbers and explosion amplitudes of all fireworks
changes over iterations by rates of 1.1 and 0.9, respectively; mutation operator is removed; and
independent selection is adopted. The only difference between the EFWA-D1 and the EFWA-D2
is that a simplified adaptive amplitude takes the place of the exponentially decreased amplitude.
Experimental results indicate that the performances of both algorithms measured by offline error
are better than that of some previous dynamic optimization techniques including hyper-mutation,
random immigrants, memory search and self-organization scouts.

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
A Comprehensive Review of the Fireworks Algorithm 121:11

3.4 Multi-objective Optimization

Liu et al. [113] proposed a multi-objective FWA based on the S-metric. There are two main modifi-
cations based on the original FWA. (1) The candidates with the largest S-metrics are selected as the
fireworks of the next generations. (2) An external archive is adopted to maintain the best solution
set. In each generation, when updating the archive, the individuals with the smallest S-metrics are
removed one by one. Experimental results on six multi-objective test problems indicate that the
proposed algorithm (S-MOFWA) outperforms NSGA-II, SPEA-2, and PESA-2.
Bejinariu et al. [21] proposed two different approaches to extending the FWA to solve multi-
objective problems: (1) scalarization, i.e., transforming a multi-objective optimization problem into
a single-objective optimization problem in which the object is computed as a weighted sum of
the objective functions, and (2) random selection of non-dominated solutions as the fireworks.
Experimental results on one multi-objective problem are shown, but there is no comparison against
other algorithms.
Chen et al. [32] proposed a hybrid multi-objective optimization algorithm based on the MOEA/D
framework [231], which is composed of an offspring generation method and two different replace-
ment strategies. In the proposed algorithm, the explosion operator and the Gaussian mutation
operator of the FWA are used for generating offspring. Experimental results on 19 test functions
show the advantage of the proposed algorithm over several other MOEA/D variants.

4 THEORETICAL ANALYSES
Liu et al. [112] conducted analyses on the convergence and the time complexity of the FWA. By
considering the FWA as a Markov stochastic process, they pointed out that the original FWA is
globally convergent (because it contains a global random mutation operator) and they also gave
estimation of its convergence time.
Using a similar procedure as in Reference [112], Li et al. [104] also gave proof of global conver-
gence of their improved FWA. While Gao et al. [52] gives another kind of proof for their algorithm,
which shows that the population’s probability density function should be closely concentrated
near the objective function’s global optimal value after sufficient iterations.
Li et al. [101] analyzed the properties of the guiding mutation operator on a (basically) two-
dimensional objective function. They pointed out that (1) the length of the guiding vector is rela-
tively short on irrelevant directions, and therefore the direction is accurate, and (2) the length of
the guiding vector is relatively long on relevant directions if the local optimum is outside the explo-
sion amplitude and short otherwise. Therefore, the step size of the mutation operator is adaptive
to different search processes.
Li and Tan [98] gave sufficient conditions of local convergence of the BBFWA. According to
their analyses, the sufficient condition can be relaxed when the amplification coefficient is adopted,
which justifies the design of the dynamic explosion amplitude. Moreover, generating more sparks
is helpful to avoiding premature convergence.

5 APPLICATIONS
So far, different versions of the FWA have been applied in a wide range of real-world problems,
which are summarized in Table 1.

5.1 Supervised Learning

He et al. [67] used the FWA to “optimize the parameters of a local-concentration model for spam
detection.” Zheng and Tan [240] used the FWA, PSO, and DE to optimize the parameters of a unified
distance measure for palmprint and finger-vein identification. Lihu and Holban [109] used the FWA

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
121:12 J. Li and Y. Tan

to maximize the Kullback-Leibler divergence between candidate motifs. Alamaniotis et al. [2, 3]
used the original FWA to optimize the coefficients of “a linear combination of known template
signature patterns” for radioisotopic identification. Gonsalves [58] used the FWA to optimize the
feature subset for software cost estimation. Ma and Niu [121] used the FWA to optimize the feature
subset for icing forecasting of high voltage transmission line. Sreeja [161] proposed a weighted
pattern matching approach for imbalanced classification in which the FWA is adopted for feature
and weight optimization. Tuba et al. [184] used the EFWA to optimize the parameters of the support
vector machine, whose performance was demonstrated on standard datasets. Duan et al. [46] used
the FWA to optimize the coefficients of the twin support vector regression prediction model for
basic oxygen steelmaking endpoint prediction. Zhang et al. [230] used the original FWA to optimize
the parameters of the multiclass support vector machine for the task of unnatural control chart
pattern recognition. Lei et al. [92] used the BBFWA to optimize the least squares support vector
machine for short-term power load forecasting. Khuat and Le [82] used the FWA to train the
feed forward neural network at the beginning stage (continued with the Levenberq-Marquardt
algorithm) for the task of agile software effort prediction. Dutta et al. [47] and Salman et al. [151]
respectively used the FWA to optimize the feed forward neural network for classification tasks
in medical data processing. Khuat et al. [83, 84] used the FWA to train the feed forward neural
network at the beginning stage (continued with the back-propagation algorithm) for the task of
stock price estimation. Suksri and Kimpan [167] used the FWA to train the feed forward neural
network for weather forecasting. Gonsalves [59] used PSO and the EFWA to train feed forward
neural networks for classification tasks. Bolaji et al. [24] used the FWA to train the feed forward
neural network for classification tasks. Zhang J. and Zhang H. [228] used the FWA to train a three-
layer neural network for predicting short-time traffic flow. Xue et al. [204–206] used the FWA to
train a linear model for classification tasks. Zalasiński et al. [224] used the FWA to train a fuzzy
system for predicting values of features describing the dynamic signature. Tao and Ye [175] used
the FWA to optimize a Gaussian process regression model for WiFi indoor positioning.

5.2 Unsupervised Learning

Yang and Tan [209] used the FWA and several other evolutionary algorithms to initialize k-means
for document clustering. Karimov and Ozbayoglu [79] proposed a hybrid algorithm of cuckoo
search and the EFWA to initialize k-means for big data clustering. Mattos et al. [125] evaluated
several meta-heuristics including the FWA for clustering of supply chain data. Tuba et al. [179]
to BBFWA to optimize the centers in k-means in the first iteration for web data clustering. Liu
et al. [110] proposed a discrete FWA to optimize the number of clusters along with the cluster-
heads in wireless sensor clustering. Bouarara et al. [25] proposed a framework using the fireworks
algorithm to optimize query-document relevance for automatic web information retrieval. Si [158]
used the fireworks algorithm for evolving computer programs automatically. Guendouz et al. [63]
proposed a discrete version of the FWA for community detection in complex networks. Ma and
Xia [122] proposed another approach for community detection based on a discrete FWA with local
double ring initialization.

5.3 Scheduling/Routing
Zheng et al. [242] proposed a multi-objective version of the FWA to optimize the dosage of fertil-
izers. Several groups of researchers [4, 68–70, 93, 130, 132, 133, 234] proposed or adopted different
variants of the FWA to minimize power loss in network reconfiguration. Abdulmajeed and Ayob
[1] and Yang and Ke [207], respectively, adapted the FWA to solve the capacitated vehicle routing
problems. Cai et al. [27] established a model of “the vehicle routing problem with multiple time
windows” and proposed a quantum fireworks evolutionary algorithm to solve it. Bacanin and Tuba

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
A Comprehensive Review of the Fireworks Algorithm 121:13

[11] adapted the FWA for constrained portfolio investment based on the extended mean-variance
model. Zhang and Liu [233] also used the FWA to optimize portfolio investment, but this is based on
the classical mean-variance model. Liu et al. [116] proposed a binary version of the FWA for multi-
satellite control resource scheduling. Tuba et al. [162, 189, 190] proposed several approaches for
RFID planning based on different versions of the FWA. Kumar et al. [136, 146–149, 153] proposed
a binary FWA to solve the thermal unit commitment problem with constraints. Yang et al. [208]
proposed a multi-objective FWA to find services in the task of data-intensive service mashup. Tuba
et al. [8, 182, 183, 186], Liu et al. [115], and Xia et al. [197] used the FWA to find optimal sensor nodes
positions that covers the area of interest maximally. Wei et al. [196] proposed a multi-objective dis-
crete FWA for charging path planning of wireless rechargeable sensor networks. Ding et al. [40]
designed a tourist recommender system and proposed a discrete FWA to solve optimize the tourist
trip. Li and Lu [102, 118] proposed a discrete FWA for assembly sequence planning. Zhang et al.
[232] and Liu et al. [117] proposed two discrete FWAs to solve the satellite link scheduling problem.
Shi et al. [154] used the FWA to solve “the load balancing problem in the software defined cloud-fog
network.” Alihodzic [5, 6] used the FWA and the GFWA, respectively, for unmanned aerial vehi-
cle path planning. While Wang et al. [193] proposed a hyper-heuristic integrating GA, PSO, BBO,
FWA, and WWO to solve the same problem. Jadoun et al. [71, 72, 135] proposed an improved FWA
to solve the economic dispatch problem. Xue et al. [203] proposed a discrete fireworks algorithm to
solve the aircraft mission planning problem. Pavão et al. [137] proposed a two level optimization
approach based on simulated annealing and a novel rocket FWA to optimize the heat exchanger
network. Ye et al. [210] introduced local search and chaotic mutation mechanisms in the FWA to
solve the warehouse-scheduling problem. Wang et al. [194] proposed a multi-objective hybrid al-
gorithm of the FWA and gravitational search to solve the economic and environmental operation
management problem. Mnif and Bouamama [129] proposed a multi-objective FWA to solve the
multimodal transportation problem. Fu et al. [50] proposed a multi-objective discrete FWA for the
stochastic flow-shop scheduling. Ting et al. [176] used the FWA for hybrid flow shop scheduling.
Bahramian-Habil et al. [13] proposed a multi-objective FWA for fault current limiter placement.
Guo et al. [66] proposed a discrete multi-objective FWA for software project scheduling. Li et al.
[95] proposed an improved FWA for multi-core processor scheduling.

5.4 Numerical Calculation

Janecek and Tan [73–76] used several evolutionary algorithms including the FWA for the initial-
ization of the nonnegative matrix factorization. Zheng et al. [239] evaluated several versions of
the FWA on the ICSI2014 numerical optimization benchmark suite. Li et al. [94] used several ver-
sions of the FWA to optimize the parameters in the chaotic system. Reddy et al. [144, 145] used
the FWA to estimate the control points for Bezier curves/surfaces fitting. Guan et al. [62] used
the FWA to optimize the nodes in the integral interval for numerical integration. Mu et al. [131]
introduced a linearly decreased dimension number strategy in the dynFWA and parallelized it to
find the optimal perturbation of a nonlinear model.

5.5 Design/Control
Gao and Diao [51] proposed a hybrid algorithm of the FWA and the culture algorithm for digital
filter design. Pholdee and Bureerat [141] compared the performances of several different evolution-
ary algorithms on the task of truss mass minimization with dynamic constraints. While Gholizadeh
and Milany [54] proposed an improved fireworks algorithm for both truss and frame structure
optimization tasks. Rajaram et al. [143] tested the firefly algorithm and the FWA on the task of se-
lective harmonic elimination in inverter output waveforms. Goswami and Chakraborty [61] tested
the FWA and cuckoo search for the parametric optimization of laser machining processes. Later,

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
121:14 J. Li and Y. Tan

Fig. 2. GFWA for JPEG quantization table optimization.

the same authors [60] tested the FWA and gravitational search for the parametric optimization of
ultrasonic machining processes. Babu et al. [10] used the FWA to optimize the parameters of the
solar photovoltaic model. Several groups of researchers [88, 163–165, 202, 213] used the FWA to
optimize the parameters of the proportional-integral-derivative controller. BouDaher and Hoorfar
[26] and Tang et al. [174] used the FWA for electromagnetic and antenna optimization. Fortes et al.
[39] used the FWA to optimize parameters of the supplementary damping controllers. Manickam
et al. [124] and Sangeetha et al. [152] used the FWA for maximum power point tracking in the
photovoltaic system. Dou et al. [45] used the AFWA for inverse analysis of concrete dams. Guer-
reiro et al. [64], Amhaimar et al. [7], and Basílio [20], respectively, used the FWA to optimize the
design of orthogonal frequency division multiplexing schemes. Yin et al. [195, 214] respectively
proposed variants of the AFWA to optimize parameters for the control of hypersonic vehicles. Gao
et al. [52] proposed a hybrid algorithm of the FWA and quantum computation to find the optimal
cooperative mechanism of energy harvesting cognitive radio. Karkalos et al. [80] used the FWA to
optimize the parameters of the constitutive material model for stainless steel. Xie et al. [199] pro-
posed a hybrid algorithm of the FWA and PSO to improve the crashworthiness of subway vehicles.
Jeronymo et al. [77] proposed an improved FWA combined with free search and opposition-based
learning for the design of a spiral inductor. Pallone et al. [134] adapted the EFWA to optimize the
ascent path of multistage launch vehicles.

5.6 Image Processing

Tuba et al. [188], Chen et al. [28, 114], and Chen et al. [34], respectively, used the FWA to optimize
the threshold values in image segmentation. Rahmani et al. [142] proposed an explosive image
perturbation approach for privacy preserving. Shi et al. [155, 156] proposed a modified FWA that
adopts different displacements on different dimensions to locate multiple cells in microscopy im-
ages. Bejinariu et al. [22] and Tuba et al. [181, 185] used the FWA for image registration. Bejinariu
et al. [23] tested PSO, cuckoo search and the FWA on the task of image fusion. Tuba et al. [178, 187]
used the GFWA and the BBFWA, respectively, to optimize the JPEG quantization table in image
compression (see Figure 2). Wang et al. [191] used the FWA to optimize the conventional neural
network for image retrieval.

5.7 Others
Ding et al. [33, 41] used several different evolutionary algorithms for seismic waveform inversion.
Ren et al. [150] proposed a velocity-based improved FWA to estimate the thermal and optical
ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
A Comprehensive Review of the Fireworks Algorithm 121:15

properties of molten salt. Crawford et al. [37] and Tuba et al. [177] used different variants of the
FWA to solve the set covering problem. Taidi et al. [169] and Luo et al. [120] proposed discrete
FWAs to solve the travelling salesman problem. Xue et al. [201] proposed an uncertain bilevel
knapsack problem and proposed a binary backward FWA for solving it. Lee [90, 91] proposed a
modified FWA for the inverse scattering of a conducting cylinder. Łapa [87] proposed a hyper-
heuristic integrating several evolutionary operators, including explosion and mutation operators
of the FWA for non-linear modeling. Later, Łapa et al. [89] proposed a fuzzy system-based approach
for non-linear modeling, in which a hybrid algorithm of the GA and the FWA is used to optimize
the structure and parameters. Xiao et al. [198] used the EFWA to locate the critical slip surface with
the minimal safety factor in slope stability analysis. Ma et al. [123] used the AFWA for minefield
attack decision. Tuba et al. [180] used the BBFWA to solve the capacitated p-median problem. Miao
et al. [126] modified the FWA to solve the problem of mobile robot odor source localization. Chen
et al. [29] compared the performances of the complete algorithm, the heuristic algorithm and the
FWA on the problem of for task-oriented satellite agent team formation.

6 IMPLEMENTATIONS
6.1 Parallelization
Ding et al. [43] proposed the the first parallel framework of the FWA implemented on GPU, which
enjoys a speedup as high as 200× with the expense that the fireworks cannot exchange information
frequently. To guarantee the quality of the solutions, they proposed an attract-repulse mutation
operator to enhance the performance, which is similar to the Gaussian mutation in the EFWA but
the distribution is uniform. Experimental results indicate that the proposed GPU-FWA not only
outruns the original FWA and PSO but also outperforms them. Later, Ding and Tan [42] further
elevated the performance by introducing the dynamic explosion amplitude and a Cauchy mutation
operator into the GPU-FWA.
Ludwig and Dawar [119] implemented the EFWA using the MapReduce platform. Experimental
results performed on different benchmark functions indicate that the EFWA achieves a better
speedup rate than PSO on the MapReduce framework, which concurs with the finding in
Reference [43].

6.2 Others
Baidoo [14] implemented a Java version of the FWA with a simple user interface.
Different versions of the FWA have also been implemented in MATLAB, C, Java, C++, Python,
and so on. Please visit http://www.cil.pku.edu.cn/research/fwa/resources/index.html for more
information.

7 DISCUSSION
Figure 3 shows the number of the literatures of the FWA in recent years. It can be seen that the FWA
has been attracting more and more research interest and has been used to solve more and more
real-world problems. Among all the literature, the number of applications is the most, followed by
algorithmic research.

7.1 Algorithmic Research

Most algorithmic studies of the FWA place their focus on single objective optimization, which
is the basis of multi-objective optimization and applications. Improving the FWA itself and
hybridizing it with other algorithms are both valuable and important as long as the performance
can be elevated. Actually, algorithmic research on the FWA has gone deeper and deeper in

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
121:16 J. Li and Y. Tan

Table 1. Applications of the FWA

Supervised spam detection [67], palmprint and finger-vein identification [240], de novo
Learning motif prediction [109], gamma-ray spectrum analysis [2, 3], feature subset
optimization [58, 121, 161], support vector machine parameters optimization
[46, 92, 184, 230], artificial neural network training [24, 47, 59, 82–84, 151, 167,
228], linear model training [204–206], prediction of values of the dynamic
signature features [224], Gaussian process regression model training [175]
Unsupervised clustering [79, 110, 125, 179, 209], web information retrieval [25], grammatical
Learning evolution [158], community detection [63, 122]
Scheduling / fertilization in oil crop production [242], power system reconfiguration [4,
Routing 68–70, 93, 130, 132, 133, 234], vehicle routing [1, 27, 207], portfolio optimization
[11, 233], multi-satellite control resource scheduling [116], RFID network
planning [162, 189, 190], thermal unit commitment [136, 146–148, 149, 153],
data-intensive service mashup [208], wireless sensor network [8, 115, 182, 183,
186, 196, 197], tourist route planning [40], assembly sequence planning [102,
118], satellite link scheduling [117, 232], load balancing for cloud-fog network
[154], UAV path planning [5, 6, 193], economic dispatch [71, 72, 135], aircraft
mission planning [203], heat exchanger networks synthesis [137], warehouse
scheduling [210], economic/environmental operation management [194],
multimodal transportation [129], flow-shop scheduling [50, 176], fault current
limiter placement [13], software project scheduling [66], multi-core processor
scheduling [95]
Numerical nonnegative matrix factorization [73–76], numerical optimization [239],
Calculation parameter estimation of chaotic systems [94], fitting of Bezier curves/surfaces
[144, 145], numerical integration [62], conditional nonlinear optimal
perturbation [131]
Design / digital filters design [51], truss/frame structure optimization [54, 141], selective
Control harmonic elimination in PWM inverter [143], laser machining process [61],
ultrasonic machining process [60], parameter extraction of two diode solar PV
model [10], PID parameter optimization [88, 163–165, 202, 213], antenna design
[26, 174], supplementary damping controller design [39], maximum power point
tracking in PV systems [124, 152], parameter identification of concrete dams
[45], critical slip surface locating [198], design of nonlinear OFDM [7, 20, 64],
control of hypersonic vehicles [195, 214], optimization of energy harvesting
cognitive radio [52], determination of Johnson-Cook material model parameters
[80], improvement of the crashworthiness of subway vehicles [199], design of a
spiral inductor [77], optimization of the ascent path of multistage launch
vehicles [134]
Image multilevel image thresholding [28, 34, 114, 188], privacy preserving through
Processing image perturbation [142], cells tracking [155, 156], image registration [22, 181,
185], image fusion [23], image compression [178, 187], image retrieval [191]
Others seismic waveform inversion [33, 41], estimation of thermal and optical
properties of molten salt [150], set covering problem [37], maximal covering
location problem [177] travelling salesman problem [120, 169], knapsack
problem [201], inverse scattering of a conducting cylinder [90, 91], nonlinear
modeling [87, 89], slope stability analysis [198], minefield attack decision [123],
capacitated p-median problem [180], mobile robot odor source localization
[126], task-oriented satellite agent team formation [29]

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
A Comprehensive Review of the Fireworks Algorithm 121:17

Fig. 3. Number of literatures of the FWA.

several different directions recently, such as improvements of the operators, novel cooperative
mechanisms among fireworks, strategies for generating elite solutions, and more thorough
utilization of the information. However, it can also be seen from this review that some of the
studies are not based on competent knowledge of the field of FWA research. For example, some
recent studies are still based on the original FWA, which is of several obvious drawbacks. It is
part of the purpose of this review to provide a full picture for new researchers and avoid wasteful
efforts. Here we provide the following suggestions to help them make a solid contribution.
(1) If you want to improve the FWA, then please choose a state-of-the-art version as the
baseline algorithm rather than the original FWA.
(2) Make sure your benchmark is wide enough and shifted. The most convenient way is to
use standard benchmarks.
(3) Try to provide hints about when your algorithm performs well, and when your algorithm
performs not well. Because there is no cure-for-all.
(4) Conduct statistical tests to show the significance of your improvement.
(5) If you are trying to design a hybrid algorithm, then make sure your hybrid algorithm
outperforms all component algorithms significantly.
Constraint handling is one of the most important topics in optimization. However, although
some discussions have been conducted on the mapping rules of the FWA (which is used to han-
dle boundary constraints), there are not enough studies on handling equality or other inequality
constraints. According to the framework of the FWA, other inequality constraints can be handled
by new kinds of mapping rules, while handling equality constraints may require new kinds of
explosion/mutation operators.
How to solve multi-objective optimization and dynamic optimization problems using the FWA
are also relatively new issues and deserve more attention in the future. Currently, multi-objective
and dynamic FWAs can outperform classic methods on standard test functions, which has prelim-
inarily shown its potential. Existing studies have also shown that designing new problem-solving
frameworks using the principles of the FWA, and integrating the FWA into existing frameworks
(like MOEA/D) are both feasible approaches.

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
121:18 J. Li and Y. Tan

Due to the unique framework of the FWA, we think its algorithmic research should be very
attractive and challenging, because it may be able to solve problems that other algorithms cannot.

7.2 Theoretical Analyses

Theoretical analyses are of great importance, because they are supposed to construct the founda-
tion of algorithmic research. However, the number of theoretical analyses of the FWA is limited due
to its complex stochastic behaviors. There is only one completely theoretical paper regarding the
convergence and time complexity of the original FWA, though theoretical analyses of other FWA
variants can also be found in the literature such as References [101, 104] and Reference [98]. Fur-
ther research is needed regarding the convergence properties (especially the convergence speed)
and the dynamics of the FWA.

7.3 Applications and Implementations

The FWA has already been successfully applied in many different fields. Recently, many new ver-
sions of the FWA are proposed along with their applications in certain fields, which is in our
opinion a very good trend. Because no algorithmic research should be conducted without usage
scenarios. However, so far most of the applications are based on early versions of the FWA. We
expect more evaluations of its improved versions on real-world problems in the future. We also
encourage applications of the FWA on large-scale tasks, especially in big data processing and anal-
yses, taking advantage of the efficiency and parallelizability of the FWA.
The FWA is of a unique framework, with competitive performance and high efficiency and
parallelizability. We believe it will continue to attract more research interest and benefit the world
in the future.

REFERENCES
[1] Noora Hani Abdulmajeed and Masri Ayob. 2014. A firework algorithm for solving capacitated vehicle routing prob-
lem. Int. J. Adv. Comput. Technol. 6, 1 (2014), 79–86.
[2] Miltiadis Alamaniotis, Chan K. Choi, and Lefteri H. Tsoukalas. 2015. Application of fireworks algorithm in gamma-
ray spectrum fitting for radioisotope identification. Int. J. Swarm Intell. Res. 6, 2 (2015), 102–125.
[3] Miltiadis Alamaniotis and Lefteri H. Tsoukalas. 2018. Assessment of gamma-ray-spectra analysis method utilizing
the fireworks algorithm for various error measures. In Critical Developments and Applications of Swarm Intelligence.
IGI Global, 155–181.
[4] Hafiz Munsub Ali, Waleed Ejaz, Daniel C. Lee, and Ismail M. Khater. 2018. Optimising the power using firework-
based evolutionary algorithms for emerging IoT applications. IET Netw. 8, 1 (2018), 15–31.
[5] Adis Alihodzic. 2016. Fireworks algorithm with new feasibility-rules in solving UAV path planning. In Proceedings
of the 2016 3rd International Conference on Soft Computing & Machine Intelligence (ISCMI’16). IEEE, 53–57.
[6] Adis Alihodzic, Damir Hasic, and Elmedin Selmanovic. 2018. An effective guided fireworks algorithm for solving
UCAV path planning problem. In Proceedings of the International Conference on Numerical Methods and Applications.
Springer, 29–38.
[7] Lahcen Amhaimar, Saida Ahyoud, Ali Elyaakoubi, Abdelmoumen Kaabal, Kamal Attari, and Adel Asselman. 2018.
PAPR reduction using fireworks search optimization algorithm in MIMO-OFDM systems. J. Electr. Comput. Eng.
(2018).
[8] Aleksandra Arsic, Milan Tuba, and Milos Jordanski. 2016. Fireworks algorithm applied to wireless sensor networks
localization problem. In Proceedings of the 2016 IEEE Congress on Evolutionary Computation (CEC’16). IEEE, 4038–
4044.
[9] Anne Auger and Nikolaus Hansen. 2005. A restart CMA evolution strategy with increasing population size. In
Proceedings of the 2005 IEEE Congress on Evolutionary Computation (CEC’05), Vol. 2. IEEE, 1769–1776.
[10] T. Sudhakar Babu, J. Prasanth Ram, K. Sangeetha, Antonino Laudani, and N. Rajasekar. 2016. Parameter extraction
of two diode solar PV model using Fireworks algorithm. Solar Energy 140 (2016), 265–276.
[11] Nebojsa Bacanin and Milan Tuba. 2015. Fireworks algorithm applied to constrained portfolio optimization problem.
In Proceedings of the 2015 IEEE Congress on Evolutionary Computation (CEC’15). IEEE, 1242–1249.

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
A Comprehensive Review of the Fireworks Algorithm 121:19

[12] Nebojsa Bacanin, Milan Tuba, and Marko Beko. 2015. Hybridized fireworks algorithm for global optimization. Math.
Methods Syst. Sci. Eng. 1 (2015), 108–114.
[13] Hossein Bahramian-Habil, Ehsan Azad-Farsani, Behrooz Vahidi, Hossein Askarian-Abyaneh, and Gevork Babamalek
Gharehpetian. 2017. Fault current limiter placement using multi-objective firework algorithm. Electr. Power Comp.
Syst. 45, 17 (2017), 1929–1940.
[14] Evans Baidoo. 2017. Fireworks algorithm for unconstrained function optimization problems. Appl. Comput. Sci. 13,
1 (2017), 61–74.
[15] Juan Barraza, Patricia Melin, Fevrier Valdez, and Claudia González. 2017. Fireworks algorithm (FWA) with adaptation
of parameters using fuzzy logic. In Nature-Inspired Design of Hybrid Intelligent Systems. Springer, 313–327.
[16] Juan Barraza, Patricia Melin, Fevrier Valdez, and Claudia I. Gonzalez. 2016. Fuzzy FWA with dynamic adaptation of
parameters. In Proceedings of the 2016 IEEE Congress on Evolutionary Computation (CEC’16). IEEE, 4053–4060.
[17] Juan Barraza, Patricia Melin, Fevrier Valdez, and Claudia I. Gonzalez. 2017. Fuzzy fireworks algorithm based on a
sparks dispersion measure. Algorithms 10, 3 (2017), 83.
[18] Juan Barraza, Patricia Melin, Fevrier Valdez, Claudia I. Gonzalez, and Oscar Castillo. 2017. Iterative fireworks al-
gorithm with fuzzy coefficients. In Proceedings of the 2017 IEEE International Conference on Fuzzy Systems (FUZZ-
IEEE’17). IEEE, 1–6.
[19] Juan Barraza, Luis Rodríguez, Oscar Castillo, Patricia Melin, and Fevrier Valdez. 2018. A new hybridization approach
between the fireworks algorithm and grey wolf optimizer algorithm. J. Optimiz. (2018).
[20] Diogo Francisco Frederique Basílio. 2018. Approaching the Optimal Performance of Nonlinear OFDM with FWA Tech-
niques. Ph.D. Dissertation.
[21] Silviou-Ioan Bejinariu, Hariton Costin, Florin Rotaru, Ramona Luca, and Cristina Niţă. 2016. Fireworks algorithm
based single and multi-objective optimization. Bull. Polytech. Inst. Iasi, Autom. Contr. Comput. Sci. Sec. 62, 66 (2016),
19–34.
[22] Silviu-Ioan Bejinariu, Hariton Costin, Florin Rotaru, Ramona Luca, Cristina Diana Niţă, and Camelia Lazăr. 2016.
Fireworks algorithm based image registration. In Proceedings of the International Workshop Soft Computing Applica-
tions. Springer, 509–523.
[23] Silviu-Ioan Bejinariu, Ramona Luca, and Hariton Costin. 2016. Nature-inspired algorithms based multispectral image
fusion. In Proceedings of the 2016 International Conference and Exposition on Electrical and Power Engineering (EPE’16).
IEEE, 010–015.
[24] Asaju La’aro Bolaji, Aminu Ali Ahmad, and Peter Bamidele Shola. 2018. Training of neural network for pattern
classification using fireworks algorithm. Int. J. Syst. Assur. Eng. Manage. 9, 1 (01 Feb. 2018), 208–215. DOI:https://
doi.org/10.1007/s13198-016-0526-z
[25] Hadj Ahmed Bouarara, Reda Mohamed Hamou, Abdelmalek Amine, and Amine Rahmani. 2015. A fireworks algo-
rithm for modern web information retrieval with visual results mining. Int. J. Swarm Intell. Res. 6, 3 (2015), 1–23.
[26] Elie BouDaher and Ahmad Hoorfar. 2016. Fireworks algorithm: A new swarm intelligence technique for electro-
magnetic optimization. In Proceedings of the 2016 IEEE International Symposium on Antennas and Propagation (AP-
SURSI’16). IEEE, 575–576.
[27] Yanguang Cai, Yuanhang Qi, Houren Chen, Hao Cai, and Ole Hejlesen. 2018. Quantum fireworks evolutionary
algorithm for vehicle routing problem in supply chain with multiple time windows. In Proceedings of the 2018 2nd
IEEE Advanced Information Management, Communicates, Electronic and Automation Control Conference (IMCEC’18).
IEEE, 383–388.
[28] Hongwei Chen, Xingpeng Deng, Laiyi Yan, and Zhiwei Ye. 2017. Multilevel thresholding selection based on the
fireworks algorithm for image segmentation. In Proceedings of the 2017 International Conference on Security, Pattern
Analysis, and Cybernetics (SPAC’17). IEEE, 175–180.
[29] Hao Chen, Shu Yang, Jun Li, and Ning Jing. 2018. Exact and heuristic methods for observing task-oriented satellite
cluster agent team formation. Math. Probl. Eng. (2018).
[30] Junfeng Chen, Qiwen Yang, Jianjun Ni, Yingjuan Xie, and Shi Cheng. 2015. An improved fireworks algorithm with
landscape information for balancing exploration and exploitation. In Proceedings of the 2015 IEEE Congress on Evo-
lutionary Computation (CEC’15). IEEE, 1272–1279.
[31] Shuangqing Chen, Yang Liu, Lixin Wei, and Bing Guan. 2018. PS-FW: A hybrid algorithm based on particle swarm
and fireworks for global optimization. Comput. Intell. Neurosci. (2018).
[32] Xiaoji Chen, Chuan Shi, Aimin Zhou, Siyong Xu, and Bin Wu. 2018. A hybrid replacement strategy for MOEA/D. In
Proceedings of the International Conference on Bio-Inspired Computing: Theories and Applications. Springer, 246–262.
[33] Yanyang Chen, Yanbin Wang, and Yuansheng Zhang. 2017. Crustal velocity structure of central Gansu Province
from regional seismic waveform inversion using firework algorithm. Earthq. Sci. 30, 2 (2017), 81–89.
[34] Yuqing Chen, Wei Yang, Minzan Li, Ziyuan Hao, Peng Zhou, and Hong Sun. 2018. Research on pest image processing
method based on Android thermal infrared lens. IFAC-PapersOnLine 51, 17 (2018), 173–178.

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
121:20 J. Li and Y. Tan

[35] Rong Cheng, Yanping Bai, Yu Zhao, Xiuhui Tan, and Ting Xu. 2019. Improved fireworks algorithm with information
exchange for function optimization. Knowledge-Based Systems 163 (2019), 82–90.
[36] Shi Cheng, Quande Qin, Junfeng Chen, Yuhui Shi, and Qingyu Zhang. 2015. Analytics on fireworks algorithm solving
problems with shifts in the decision space and objective space. Int. J. Swarm Intell. Res. 6, 2 (2015), 52–86.
[37] Broderick Crawford, Ricardo Soto, Gonzalo Astudillo, Eduardo Olguín, and Sanjay Misra. 2016. Solving set covering
problem with fireworks explosion. In Proceedings of the International Conference on Computational Science and Its
Applications. Springer, 273–283.
[38] Swagatam Das and P. N. Suganthan. 2010. Problem definitions and evaluation criteria for CEC 2011 competition on
testing evolutionary algorithms on real world optimization problems. (2010).
[39] Elenilson de Vargas Fortes, Leonardo H. Macedo, Luís Fabiano Barone Martins, and Ednei Luiz Miotto. 2017. A
Fireworks Metaheuristic for the Design of PSS and TCSC-POD Controllers for Small-Signal Stability Studies.
[40] Hui Ding, Liangjun Ke, and Zihao Geng. 2016. Route planning in a new tourist recommender system: A fireworks
algorithm based approach. In Proceedings of the 2016 IEEE Congress on Evolutionary Computation (CEC’16). IEEE,
4022–4028.
[41] Ke Ding, Yanyang Chen, Yanbin Wang, and Ying Tan. 2015. Regional seismic waveform inversion using swarm
intelligence algorithms. In Proceedings of the 2015 IEEE Congress on Evolutionary Computation (CEC’15). IEEE, 1235–
1241.
[42] Ke Ding and Ying Tan. 2015. Attract-repulse fireworks algorithm and its CUDA implementation using dynamic
parallelism. Int. J. Swarm Intell. Res. 6, 2 (2015), 1–31.
[43] Ke Ding, Shaoqiu Zheng, and Ying Tan. 2013. A gpu-based parallel fireworks algorithm for optimization. In Proceed-
ings of the 15th Annual Conference on Genetic and Evolutionary Computation. ACM, 9–16.
[44] Marco Dorigo, Mauro Birattari, and Thomas Stutzle. 2006. Ant colony optimization. IEEE Comput. Intell. Mag. 1, 4
(2006), 28–39.
[45] Siqi Dou, Junjie Li, and Fei Kang. 2017. Parameter identification of concrete dams using swarm intelligence algo-
rithm. Eng. Comput. 34, 7 (2017), 2358–2378.
[46] Jiaojiao Duan, Qiang Qu, Chuang Gao, and Xuebo Chen. 2017. BOF steelmaking endpoint prediction based on FWA-
TSVR. In Proceedings of the 2017 36th Chinese Control Conference (CCC’17). IEEE, 4507–4511.
[47] Ram Kinkar Dutta, Nabin Kanti Karmakar, and Tapas Si. 2016. Artificial neural network training using fireworks
algorithm in medical data mining. Int. J. Comput. Appl. 137, 1 (2016), 1–5.
[48] Muzaffar Eusuff, Kevin Lansey, and Fayzul Pasha. 2006. Shuffled frog-leaping algorithm: A memetic meta-heuristic
for discrete optimization. Eng. Optimiz. 38, 2 (2006), 129–154.
[49] Pushpa Farswan and Jagdish Chand Bansal. 2018. Fireworks-inspired biogeography-based optimization. Soft Com-
put. 23, 16 (2018), 7091–7115.
[50] Yaping Fu, Jinliang Ding, Hongfeng Wang, and Junwei Wang. 2018. Two-objective stochastic flow-shop scheduling
with deteriorating and learning effect in Industry 4.0-based manufacturing system. Appl. Soft Comput. 68 (2018),
847–855.
[51] Hongyuan Gao and Ming Diao. 2011. Cultural firework algorithm and its application for digital filters design. Int. J.
Model. Ident. Contr. 14, 4 (2011), 324–331.
[52] Hongyuan Gao, Yanan Du, and Chenwan Li. 2018. Quantum fireworks algorithm for optimal cooperation mechanism
of energy harvesting cognitive radio. J. Syst. Eng. Electr. 29, 1 (2018), 18–30.
[53] Hongyuan Gao and Chenwan Li. 2015. Opposition-based quantum firework algorithm for continuous optimisation
problems. Int. J. Comput. Sci. Math. 6, 3 (2015), 256–265.
[54] Saeed Gholizadeh and Arman Milany. 2018. An improved fireworks algorithm for discrete sizing optimization of
steel skeletal structures. Eng. Optimiz. 50, 11 (2018), 1829–1849. DOI:https://doi.org/10.1080/0305215X.2017.1417402
arXiv:https://doi.org/10.1080/0305215X.2017.1417402
[55] David E. Goldberg and John H. Holland. 1988. Genetic algorithms and machine learning. Mach. Learn. 3, 2 (01 Oct.
1988), 95–99. DOI:https://doi.org/10.1023/A:1022602019183
[56] Chibing Gong. 2016. Chaotic adaptive fireworks algorithm. In Proceedings of the International Conference in Swarm
Intelligence. Springer, 515–525.
[57] Chibing Gong. 2016. Opposition-based adaptive fireworks algorithm. Algorithms 9, 3 (2016), 43.
[58] Tad Gonsalves. 2015. Feature subset optimization through the fireworks algorithm. Int. J. Electr. Comput. Sci. Eng. 4,
3 (2015), 211–218.
[59] Tad Gonsalves. 2016. Two diverse swarm intelligence techniques for supervised learning. In Psychology and Mental
Health: Concepts, Methodologies, Tools, and Applications. IGI Global, 849–861.
[60] Debkalpa Goswami and Shankar Chakraborty. 2015. Parametric optimization of ultrasonic machining process using
gravitational search and fireworks algorithms. Ain Shams Eng. J. 6, 1 (2015), 315–331.

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
A Comprehensive Review of the Fireworks Algorithm 121:21

[61] D. Goswami and S. Chakraborty. 2015. A study on the optimization performance of fireworks and cuckoo search
algorithms in laser machining processes. J. Inst. Eng. Ser. C 96, 3 (2015), 215–229.
[62] Jun-xia Guan, Qing-hua Su, Wang Li, and Chen Yu. 2016. A numerical integration method based on fireworks algo-
rithm. DEStech Trans. Comput. Sci. Eng. (2016), 190–195.
[63] Mohamed Guendouz, Abdelmalek Amine, and Reda Mohamed Hamou. 2017. A discrete modified fireworks algo-
rithm for community detection in complex networks. Appl. Intell. 46, 2 (2017), 373–385.
[64] Joao Guerreiro, Marko Beko, Rui Dinis, and Paulo Montezuma. 2017. Using the fireworks algorithm for ML detection
of nonlinear OFDM. In Proceedings of the 2017 IEEE 86th Vehicular Technology Conference (VTC-Fall’17). IEEE, 1–5.
[65] Jinglei Guo and Wei Liu. 2018. Enhanced fireworks algorithm with an improved gaussian sparks operator. In Pro-
ceedings of the International Symposium on Intelligence Computation and Applications. Springer, 38–49.
[66] Yinan Guo, Jianjiao Ji, Junhua Ji, Dunwei Gong, Jian Cheng, and Xiaoning Shen. 2019. Firework-based software
project scheduling method considering the learning and forgetting effect. Soft Comput. 23, 13 (2019), 5019–5034.
[67] Wenrui He, Guyue Mi, and Ying Tan. 2013. Parameter optimization of local-concentration model for spam detection
by using fireworks algorithm. In Proceedings of the International Conference in Swarm Intelligence. Springer, 439–450.
[68] Hiromitsu Ikegami and Hiroyuki Mori. 2018. Development of discrete CoFFWA for distribution network reconfigu-
rations. Electr. Eng. Jpn. 205, 3 (2018), 55–62.
[69] A. Mohamed Imran and M. Kowsalya. 2014. A new power system reconfiguration scheme for power loss minimiza-
tion and voltage profile enhancement using fireworks algorithm. Int. J. Electr. Power Ener. Syst. 62 (2014), 312–322.
[70] A. Mohamed Imran, M. Kowsalya, and D. P. Kothari. 2014. A novel integration technique for optimal network re-
configuration and distributed generation placement in power distribution networks. Int. J. Electr. Power Ener. Syst.
63 (2014), 461–472.
[71] Vinay Kumar Jadoun, Vipin Chandra Pandey, Nikhil Gupta, Khaleequr Rehman Niazi, and Anil Swarnkar. 2018.
Integration of renewable energy sources in dynamic economic load dispatch problem using an improved fireworks
algorithm. IET Renew. Power Gener. 12, 9 (2018), 1004–1011.
[72] Vinay K. Jadoun, Mukesh K. Shah, Vipin C. Pandey, Nikhil Gupta, Khaleequr R. Niazi, and Anil Swarnkar. 2016.
Multi-area dynamic economic dispatch problem with multiple fuels using improved fireworks algorithm. Int. J.
Electr. Ener. 4, 4 (2016), 204–208.
[73] Andreas Janecek and Ying Tan. 2011. Iterative improvement of the multiplicative update NMF algorithm us-
ing nature-inspired optimization. In Proceedings of the 2011 7th International Conference on Natural Computation
(ICNC’11), Vol. 3. IEEE, 1668–1672.
[74] Andreas Janecek and Ying Tan. 2011. Swarm intelligence for non-negative matrix factorization. Int. Jo. Swarm Intell.
Res. 2, 4 (2011), 12–34.
[75] Andreas Janecek and Ying Tan. 2011. Using population based algorithms for initializing nonnegative matrix factor-
ization. In Proceedings of the International Conference in Swarm Intelligence. Springer, 307–316.
[76] Andreas Janecek and Ying Tan. 2017. Swarm intelligence for dimensionality reduction: How to improve the non-
negative matrix factorization with nature-inspired optimization methods. In Nature-Inspired Computing: Concepts,
Methodologies, Tools, and Applications. IGI Global, 1564–1589.
[77] Daniel Cavalcanti Jeronymo, Jean Vianei Leite, Viviana Cocco Mariani, Leandro dos Santos Coelho, and Sotirios K.
Goudos. 2018. Spiral inductor design based on fireworks optimization combined with free search. In Proceedings of
the 7th International Conference on Modern Circuits and Systems Technologies (MOCAST’18). IEEE, 1–4.
[78] Dervis Karaboga and Bahriye Basturk. 2007. A powerful and efficient algorithm for numerical function optimization:
Artificial bee colony (ABC) algorithm. J. Glob. Optimiz. 39, 3 (2007), 459–471.
[79] Jeyhun Karimov and Murat Ozbayoglu. 2015. High quality clustering of big data and solving empty-clustering prob-
lem with an evolutionary hybrid algorithm. In Proceedings of the 2015 IEEE International Conference on Big Data (Big
Data’15). IEEE, 1473–1478.
[80] Nikolaos E. Karkalos and Angelos P. Markopoulos. 2018. Determination of Johnson-Cook material model parameters
by an optimization approach using the fireworks algorithm. Proc. Manufact. 22 (2018), 107–113.
[81] J. Kennedy and R. Eberhart. 1995. Particle swarm optimization. In Proceedings of the IEEE International Conference
on Neural Networks, 1995, Vol. 4. IEEE, 1942–1948.
[82] Thanh Tung Khuat and My Hanh Le. 2016. An effort estimation approach for agile software development using
fireworks algorithm optimized neural network. Int. J. Comput. Sci. Inf. Secur. 14 (2016), 122–130.
[83] Thanh Tung Khuat and My Hanh Le. 2017. An application of artificial neural networks and fuzzy logic on the stock
price prediction problem. Int. J. Inf. Vis. 1, 2 (2017), 40–49.
[84] Thanh Tung Khuat, Thi Bich Loan Nguyen, Quang Chanh Le, and Thi My Hanh Le. 2016. Applying artificial neural
network optimized by fireworks algorithm for stock price estimation. ICTACT J. Soft Comput. 6, 3 (2016), 1183–1191.
[85] Vijay Kumar, Jitender Kumar Chhabra, and Dinesh Kumar. 2015. Optimal choice of parameters for fireworks algo-
rithm. Proc. Comput. Sci. 70 (2015), 334–340.

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
121:22 J. Li and Y. Tan

[86] Ibai Lana, Javier Del Ser, and Manuel Vélez. 2017. A novel Fireworks Algorithm with wind inertia dynamics and
its application to traffic forecasting. In Proceedings of the 2017 IEEE Congress on Evolutionary Computation (CEC’17).
IEEE, 706–713.
[87] Krystian Łapa. 2017. Population-based algorithm with selectable evolutionary operators for nonlinear modeling. In
Proceedings of the International Conference on Information Systems Architecture and Technology. Springer, 15–26.
[88] Krystian Łapa and Krzysztof Cpałka. 2016. On the application of a hybrid genetic-firework algorithm for controllers
structure and parameters selection. In Information Systems Architecture and Technology: Proceedings of the 36th In-
ternational Conference on Information Systems Architecture and Technology–ISAT 2015–Part I. Springer, 111–123.
[89] Krystian Łapa, Krzysztof Cpałka, and Leszek Rutkowski. 2018. New aspects of interpretability of fuzzy systems for
nonlinear modeling. In Advances in Data Analysis with Computational Intelligence Methods. Springer, 225–264.
[90] Kun-Chou Lee. 2017. Inverse scattering of a conducting cylinder in free space by modified fireworks algorithm.
Progr. Electromagn. Res. 59 (2017), 135–146.
[91] Kun-Chou Lee. 2018. Microwave imaging of a conducting cylinder buried in a lossless half space by modified fire-
works algorithm. Microw. Opt. Technol. Lett. 60, 6 (2018), 1374–1381.
[92] Caijia Lei, Binghua Fang, Hui Gao, Wei Jia, and Wei Pan. 2018. Short-term power load forecasting based on Least
Squares Support Vector Machine optimized by Bare Bones Fireworks algorithm. In Proceedings of the 2018 IEEE 3rd
Advanced Information Technology, Electronic and Automation Control Conference (IAEAC’18). IEEE, 2231–2235.
[93] K. Lenin, B. Ravindhranath Reddy, and M. Surya Kalavathi. 2015. Reduction of real power loss by upgraded fireworks
algorithm. Int. J. Adv. Eng. Sci. 4, 2 (2015), 31–44.
[94] Hao Li, Peng Bai, Jun-Jie Xue, Jie Zhu, and Hui Zhang. 2015. Parameter estimation of chaotic systems using fireworks
algorithm. In International Conference in Swarm Intelligence. Springer, 457–467.
[95] Jingmei Li, Lanting Li, Jiaxiang Wang, and He Li. 2018. A CMP thread scheduling strategy based on improved
firework algorithm. In Proceedings of the 2018 4th International Conference on Computational Intelligence & Commu-
nication Technology (CICT’18). IEEE, 1–6.
[96] Junzhi Li and Ying Tan. 2015. Orienting mutation based fireworks algorithm. In Proceedings of the 2015 IEEE Congress
on Evolutionary Computation (CEC’15). IEEE, 1265–1271.
[97] Junzhi Li and Ying Tan. 2016. Enhancing interaction in the fireworks algorithm by dynamic resource allocation and
fitness-based crowdedness-avoiding strategy. In Proceedings of the 2016 IEEE Congress on Evolutionary Computation
(CEC’16). IEEE, 4015–4021.
[98] Junzhi Li and Ying Tan. 2018. The bare bones fireworks algorithm: A minimalist global optimizer. Appl. Soft Comput.
62 (2018), 454–462.
[99] J. Li and Y. Tan. 2018. Loser-out tournament-based fireworks algorithm for multimodal function optimization. IEEE
Trans. Evol. Comput. 22, 5 (Oct. 2018), 679–691. DOI:https://doi.org/10.1109/TEVC.2017.2787042
[100] Junzhi Li, Shaoqiu Zheng, and Ying Tan. 2014. Adaptive fireworks algorithm. In Proceedings of the 2014 IEEE Congress
on Evolutionary Computation (CEC’14). IEEE, 3214–3221.
[101] Junzhi Li, Shaoqiu Zheng, and Ying Tan. 2017. The effect of information utilization: Introducing a novel guiding
spark in the fireworks algorithm. IEEE Trans. Evol. Comput. 21, 1 (2017), 153–166.
[102] Jun-Ying Li and Cong Lu. 2016. Assembly sequence planning with fireworks algorithm. Int. J. Model. Optimiz. 6, 3
(2016), 195–198.
[103] Xiguang Li, Shoufei Han, Liang Zhao, and Changqing Gong. 2017. Adaptive fireworks algorithm based on two-
master sub-population and new selection strategy. In Proceedings of the International Conference on Neural Informa-
tion Processing. Springer, 70–79.
[104] Xi-Guang Li, Shou-Fei Han, and Chang-Qing Gong. 2017. Analysis and improvement of fireworks algorithm. Algo-
rithms 10, 1 (2017), 26.
[105] Xi-Guang Li, Shou-Fei Han, Liang Zhao, Chang-Qing Gong, and Xiao-Jing Liu. 2017. Adaptive mutation dynamic
search fireworks algorithm. Algorithms 10, 2 (2017), 48.
[106] J. J. Liang, B. Y. Qu, and P. N. Suganthan. 2013. Problem definitions and evaluation criteria for the CEC 2014 special
session and competition on single objective real-parameter numerical optimization. (2013).
[107] J. J. Liang, B. Y. Qu, P. N. Suganthan, and Q. Chen. 2014. Problem definitions and evaluation criteria for the CEC
2015 competition on learning-based real-parameter single objective optimization. (2014).
[108] J. J. Liang, B. Y. Qu, P. N. Suganthan, and Alfredo G. Hernández-Díaz. 2013. Problem definitions and evaluation
criteria for the CEC 2013 special session on real-parameter optimization. (2013).
[109] Andrei Lihu and Ştefan Holban. 2015. De novo motif prediction using the fireworks algorithm. Int. J. Swarm Intell.
Res. 6, 3 (2015), 24–40.
[110] Feng-Zeng Liu, Bing Xiao, Hao Li, and Li Cai. 2018. Discrete fireworks algorithm for clustering in wireless sensor
networks. In Proceedings of the International Conference on Swarm Intelligence. Springer, 273–282.

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
A Comprehensive Review of the Fireworks Algorithm 121:23

[111] Jianhua Liu, Shaoqiu Zheng, and Ying Tan. 2013. The improvement on controlling exploration and exploitation of
firework algorithm. In Proceedings of the International Conference in Swarm Intelligence. Springer, 11–23.
[112] Jianhua Liu, Shaoqiu Zheng, and Ying Tan. 2014. Analysis on global convergence and time complexity of fireworks
algorithm. In Proceedings of the 2014 IEEE Congress on Evolutionary Computation (CEC’14). IEEE, 3207–3213.
[113] Lang Liu, Shaoqiu Zheng, and Ying Tan. 2015. S-metric based multi-objective fireworks algorithm. In Proceedings of
the 2015 IEEE Congress on Evolutionary Computation (CEC’15). IEEE, 1257–1264.
[114] Wei Liu, Heng Shi, Xingyang He, Shang Pan, Zhiwei Ye, and Yingbin Wang. 2018. An application of optimized Otsu
multi-threshold segmentation based on fireworks algorithm in cement SEM image. J. Algor. Comput. Technol. (2018).
[115] Xiaolei Liu, Xiaosong Zhang, and Qingxin Zhu. 2017. Enhanced fireworks algorithm for dynamic deployment of
wireless sensor networks. In Proceedings of the 2017 2nd International Conference on Frontiers of Sensors Technologies
(ICFST’17). IEEE, 161–165.
[116] Zhenbao Liu, Zuren Feng, and Liangjun Ke. 2015. Fireworks algorithm for the multi-satellite control resource sched-
uling problem. In Proceedings of the 2015 IEEE Congress on Evolutionary Computation (CEC’15). IEEE, 1280–1286.
[117] Zhenbao Liu, Zuren Feng, and Liangjun Ke. 2016. A modified fireworks algorithm for the multi-resource range
scheduling problem. In Proceedings of the International Conference in Swarm Intelligence. Springer, 535–543.
[118] Cong Lu and Jun-Ying Li. 2017. Assembly sequence planning considering the effect of assembly resources with a
discrete fireworks algorithm. Int. J. Adv. Manufact. Technol. 93, 9–12 (2017), 3297–3314.
[119] Simone A. Ludwig and Deepak Dawar. 2015. Parallelization of enhanced firework algorithm using MapReduce. Int.
J. Swarm Intell. Res. 6, 2 (2015), 32–51.
[120] Haoran Luo, Weidi Xu, and Ying Tan. 2018. A discrete fireworks algorithm for solving large-scale travel salesman
problem. In Proceedings of the 2018 IEEE Congress on Evolutionary Computation (CEC’18). IEEE, 1–8.
[121] Tiannan Ma and Dongxiao Niu. 2016. Icing forecasting of high voltage transmission line using weighted least square
support vector machine with fireworks algorithm for feature selection. Appl. Sci. 6, 12 (2016), 438.
[122] TianRen Ma and Zhengyou Xia. 2017. A community detection algorithm based on local double rings and fireworks
algorithm. In Proceedings of the International Conference on Intelligent Data Engineering and Automated Learning.
Springer, 129–135.
[123] Yan Ma, Handong Zhao, and Wei Zhang. 2019. Research on intelligent minefield attack decision based on adaptive
fireworks algorithm. Arab. J. Sci. Eng. 44, 3 (2019), 2487–2496.
[124] Chakkarapani Manickam, Guru Praanesh Raman, Guru Raghav Raman, Saravana Ilango Ganesan, and Nagamani
Chilakapati. 2017. Fireworks enriched P&O algorithm for GMPPT and detection of partial shading in PV systems.
IEEE Trans. Power Electr. 32, 6 (2017), 4432–4443.
[125] Cesar L. C. Mattos, Guilherme A. Barreto, Dennis Horstkemper, and Bernd Hellingrath. 2017. Metaheuristic op-
timization for automatic clustering of customer-oriented supply chain data. In Proceedings of the 2017 12th Inter-
national Workshop on Self-Organizing Maps and Learning Vector Quantization, Clustering and Data Visualization
(WSOM’17). IEEE, 1–8.
[126] Yanzi Miao, Xiaoping Ma, Xin Jin, and Huijuan Lu. 2018. Mobile robot odor source localization based on modi-
fied FWA. In Proceedings of the 2018 IEEE 8th Annual International Conference on CYBER Technology in Automation,
Control, and Intelligent Systems (CYBER’18). IEEE, 854–860.
[127] Zbigniew Michalewicz. 1996. Evolution Strategies and Other Methods. Springer, Berlin, 159–177. DOI:https://doi.org/
10.1007/978-3-662-03315-9_9
[128] Seyedali Mirjalili, Seyed Mohammad Mirjalili, and Andrew Lewis. 2014. Grey wolf optimizer. Adv. Eng. Softw. 69
(2014), 46–61.
[129] Mouna Mnif and Sadok Bouamama. 2017. Firework algorithm for multi-objective optimization of a multimodal
transportation network problem. Proc. Comput. Sci. 112 (2017), 1670–1682.
[130] Hiroyuki Mori and Hiromitsu Ikegami. 2017. An advanced fireworks algorithm for distribution network reconfigu-
rations. In Proceedings of the 2017 IEEE Power & Energy Society General Meeting. IEEE, 1–5.
[131] Bin Mu, Junhui Yan, Shijin Yuan, and Jinghao Yan. 2017. Parallel dynamic search fireworks algorithm with linearly
decreased dimension number strategy for solving conditional nonlinear optimal perturbation. In Proceedings of the
2017 International Joint Conference on Neural Networks (IJCNN’17). IEEE, 2314–2321.
[132] Bogdan Constantin Neagu, Ovidiu Ivanov, and Mihai Gavrilaş. 2017. A comprehensive solution for optimal capacitor
allocation problem in real distribution networks. In Proceedings of the 2017 International Conference on Electrome-
chanical and Power Systems (SIELMEN’17). IEEE, 565–570.
[133] Shengsuo Niu, Yi Ding, and Zhirui Liang. 2015. Study on distribution network reconfiguration with various DGs. In
Proceedings of the International Conference on Materials Engineering and Information Technology Applications. Atlantis
Press, 766–771.
[134] Marco Pallone, Mauro Pontani, and Paolo Teofilatto. 2018. Performance evaluation methodology for multistage
launch vehicles with high-fidelity modeling. Acta Astronaut. 151 (2018), 522–531.

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
121:24 J. Li and Y. Tan

[135] Vipin Chandra Pandey, Vinay Kumar Jadoun, Nikhil Gupta, K. R. Niazi, and Anil Swarnkar. 2018. Improved fireworks
algorithm with chaotic sequence operator for large-scale non-convex economic load dispatch problem. Arab. J. Sci.
Eng. 43, 6 (01 Jun. 2018), 2919–2929.
[136] Lokesh Kumar Panwar, Srikanth Reddy, and Rajesh Kumar. 2015. Binary fireworks algorithm based thermal unit
commitment. Int. J. Swarm Intell. Res. 6, 2 (2015), 87–101.
[137] Leandro Vitor Pavão, Caliane Bastos Borba Costa, Mauro Antonio da Silva Sá Ravagnani, and Laureano Jiménez.
2017. Large-scale heat exchanger networks synthesis using simulated annealing and the novel rocket fireworks
optimization. AIChE J. 63, 5 (2017), 1582–1601.
[138] Yan Pei, Shaoqiu Zheng, Ying Tan, and Hideyuki Takagi. 2012. An empirical study on influence of approximation
approaches on enhancing fireworks algorithm. In Proceedings of the 2012 IEEE International Conference on Systems,
Man, and Cybernetics (SMC’12). IEEE, 1322–1327.
[139] Yan Pei, Shaoqiu Zheng, Ying Tan, and Hideyuki Takagi. 2015. Effectiveness of approximation strategy in surrogate-
assisted fireworks algorithm. Int. J. Mach. Learn. Cybernet. 6, 5 (2015), 795–810.
[140] Hakan Pekdemir and Haluk Rahmi Topcuoglu. 2016. Enhancing fireworks algorithms for dynamic optimization
problems. In Proceedings of the 2016 IEEE Congress on Evolutionary Computation (CEC’16). IEEE, 4045–4052.
[141] Nantiwat Pholdee and Sujin Bureerat. 2014. Comparative performance of meta-heuristic algorithms for mass min-
imisation of trusses with dynamic constraints. Adv. Eng. Softw. 75 (2014), 1–13.
[142] Amine Rahmani, Abdelmalek Amine, Reda Mohamed Hamou, Mohamed Elhadi Rahmani, and Hadj Ahmed
Bouarara. 2015. Privacy preserving through fireworks algorithm based model for image perturbation in big data.
Int. J. Swarm Intell. Res. 6, 3 (2015), 41–58.
[143] R. Rajaram, K. Palanisamy, Sudha Ramasamy, and Prabhu Ramanathan. 2014. Selective harmonic elimination in
PWM inverter using fire fly and fire works algorithm. Int. J. Innov. Res. Adv. Eng. 1 (2014), 55–62.
[144] K. Srinivas Reddy, Amitava Mandal, Krishna Kumar Verma, and G. Rajamohan. 2016. Fitting of Bezier curves using
the fireworks algorithm. Int. J. Adv. Eng. Technol. 9, 3 (2016), 396–403.
[145] K. Srinivas Reddy, Amitava Mandal, Krishna Kumar Verma, and G. Rajamohan. 2016. Fitting of Bezier surfaces using
the fireworks algorithm. Int. J. Adv. Eng. Technol. 9, 3 (2016), 421–429.
[146] K. Srikanth Reddy, Lokesh Kumar Panwar, Rajesh Kumar, and B. K. Panigrahi. 2016. Binary fireworks algorithm for
profit based unit commitment (PBUC) problem. Int. J. Electr. Power Ener. Syst. 83 (2016), 270–282.
[147] K. Srikanth Reddy, Lokesh Kumar Panwar, Rajesh Kumar, and Bijaya Ketan Panigrahi. 2016. Distributed resource
scheduling in smart grid with electric vehicle deployment using fireworks algorithm. J. Mod. Power Syst. Clean Ener.
4, 2 (2016), 188–199.
[148] Srikanth Reddy, Lokesh Kumar Panwar, B. K. Panigrahi, and Rajesh Kumar. 2016. Optimal demand response allo-
cation in resource scheduling with renewable energy penetration. In Proceedings of the 2016 IEEE 6th International
Conference on Power Systems (ICPS’16). IEEE, 1–6.
[149] K. Srikanth Reddy, Lokesh Panwar, B. K. Panigrahi, and Rajesh Kumar. 2018. Low carbon unit commitment (LCUC)
with post carbon capture and storage (CCS) technology considering resource sensitivity. J. Clean. Prod. 200 (2018),
161–173.
[150] Ya-Tao Ren, Hong Qi, Ming-Jian He, Shi-Ting Ruan, Li-Ming Ruan, and He-Ping Tan. 2016. Application of an im-
proved firework algorithm for simultaneous estimation of temperature-dependent thermal and optical properties of
molten salt. Int. Commun. Heat Mass Transf. 77 (2016), 33–42.
[151] Ihsan Salman, Osman N. Ucan, Oguz Bayat, and Khalid Shaker. 2018. Impact of metaheuristic iteration on artificial
neural network structure in medical data. Processes 6, 5 (2018), 57.
[152] K. Sangeetha, T. Sudhakar Babu, and N. Rajasekar. 2016. Fireworks algorithm-based maximum power point track-
ing for uniform irradiation as well as under partial shading condition. In Artificial Intelligence and Evolutionary
Computations in Engineering Systems. Springer, 79–88.
[153] B. Saravanan, C. Kumar, and D. P. Kothari. 2016. A solution to unit commitment problem using fire works algorithm.
Int. J. Electr. Power Ener. Syst. 77 (2016), 221–227.
[154] Chenhua Shi, Zhiyuan Ren, and Xiuli He. 2016. Research on load balancing for software defined cloud-fog network
in real-time mobile face recognition. In Proceedings of the International Conference on Communicatins and Networking
in China. Springer, 121–131.
[155] Jian Shi, Benlian Xu, Peiyi Zhu, and Mingli Lu. 2016. Multi-task firework algorithm for cell tracking and contour
estimation. In Proceedings of the 2016 International Conference on Control, Automation and Information Sciences (IC-
CAIS’16). IEEE, 27–31.
[156] Jian Shi, Benlian Xu, Peiyi Zhu, Mingli Lu, Wenmin Zhang, Lin Xu, and Jun Zhang. 2015. Multiple cells tracking
by firework algorithm. In Proceedings of the 2015 International Conference on Control, Automation and Information
Sciences (ICCAIS’15). IEEE, 508–511.
[157] Yuhui Shi. 2014. Developmental swarm intelligence: Developmental learning perspective of swarm intelligence al-
gorithms. Int. J. Swarm Intell. Res. 5, 1 (2014), 36–54.

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
A Comprehensive Review of the Fireworks Algorithm 121:25

[158] Tapas Si. 2016. Grammatical evolution using fireworks algorithm. In Proceedings of the 5th International Conference
on Soft Computing for Problem Solving. Springer, 43–55.
[159] Tapas Si and Rituparna Ghosh. 2015. Explosion sparks generation using adaptive transfer function in firework algo-
rithm. In Proceedings of the 2015 3rd International Conference on Signal Processing, Communication and Networking
(ICSCN’15). IEEE, 1–9.
[160] Dan Simon. 2008. Biogeography-based optimization. IEEE Trans. Evol. Comput. 12, 6 (2008), 702–713.
[161] N. K. Sreeja. 2019. A weighted pattern matching approach for classification of imbalanced data with a fireworks-
based algorithm for feature selection. Connect. Sci. 31, 2 (2019), 143–168.
[162] Ivana Strumberger, Eva Tuba, Nebojsa Bacanin, Marko Beko, and Milan Tuba. 2018. Bare bones fireworks algorithm
for the RFID network planning problem. In Proceedings of the 2018 IEEE Congress on Evolutionary Computation
(CEC’18). IEEE, 1–8.
[163] Te-Jen Su, Tsung-Ying Li, Shih-Mine Wang, Van-Manh Hoang, and Yi-Feng Chen. 2016. A novel method for con-
troller design in engineering education. World Trans. Eng. Technol. Educ. 14, 2 (2016), 288–294.
[164] Te-Jen Su, Tung-Yeh Tsou, Shih-Mine Wang, Van-Manh Hoang, and Kuo-Win Pin. 2016. A hybrid control design of
FOPID and FWA for inverted pendulum systems. J. Appl. Phys. Sci. 2, 3 (2016), 89–95.
[165] Te-Jen Su, Shih-Ming Wang, Tsung-Ying Li, Sung-Tsun Shih, and Van-Manh Hoang. 2017. Design of hybrid sliding
mode controller based on fireworks algorithm for nonlinear inverted pendulum systems. Adv. Mech. Eng. 9, 1 (2017),
1–13.
[166] P. N. Suganthan, N. Hansen, J. J. Liang, K. Deb, Y.-P. Chen, A. Auger, and S. Tiwari. 2005. Problem definitions and
evaluation criteria for the CEC 2005 special session on real-parameter optimization. (2005).
[167] Saktaya Suksri and Warangkhana Kimpan. 2016. Neural network training model for weather forecasting using fire-
works algorithm. In Proceedings of the 2016 International Computer Science and Engineering Conference (ICSEC’16).
IEEE, 1–7.
[168] Yu-Feng Sun, Jie-Sheng Wang, and Jiang-Di Song. 2016. An improved fireworks algorithm based on grouping strat-
egy of the shuffled frog leaping algorithm to solve function optimization problems. Algorithms 9, 2 (2016), 23.
[169] Zoubair Taidi, Lamia Benameur, and Jihane Alami Chentoufi. 2017. A fireworks algorithm for solving travelling
salesman problem. Int. J. Comput. Syst. Eng. 3, 3 (2017), 157–162.
[170] Ying Tan. 2015. Fireworks Algorithm: A Novel Swarm Intelligence Optimization Method. Springer.
[171] Ying Tan, Chao Yu, Shaoqiu Zheng, and Ke Ding. 2013. Introduction to fireworks algorithm. Int. J. Swarm Intell. Res.
4, 4 (2013), 39–70.
[172] Ying Tan and Yuanchun Zhu. 2010. Fireworks algorithm for optimization. In Proceedings of the International Confer-
ence in Swarm Intelligence. Springer, 355–364.
[173] Ke Tang, Xiaodong Li, P. N. Suganthan, Zhenyu Yang, and Thomas Weise. 2010. Benchmark functions for the CEC
2010 special session and competition on large-scale global optimization. (2010).
[174] Peiwen Tang, Liang Lang, Fei Hu, and Dong Zhu. 2017. The design of two-fold redundancy linear arrays in aper-
ture synthesis radiometers. In Proceedings of the 2017 IEEE International Symposium on Antennas and Propagation &
USNC/URSI National Radio Science Meeting. IEEE, 149–150.
[175] Ye Tao and Long Zhao. 2018. A novel system for WiFi radio map automatic adaptation and indoor positioning. IEEE
Trans. Vehic. Technol. 67, 11 (2018), 10683–10692.
[176] Dong Xiao Ting, Liu Chang Ming, and Han Zhong Hua. 2018. Fireworks explosion algorithm for hybrid flow shop
scheduling and optimization problem1. In IOP Conference Series: Materials Science and Engineering, Vol. 382. IOP
Publishing, 032005.
[177] Eva Tuba, Edin Dolicanin, and Milan Tuba. 2017. Guided fireworks algorithm applied to the maximal covering
location problem. In Proceedings of the International Conference in Swarm Intelligence. Springer, 501–508.
[178] Eva Tuba, Raka Jovanovic, Marko Beko, Antonio J. Tallón-Ballesteros, and Milan Tuba. 2018. Bare bones fireworks
algorithm for medical image compression. In Proceedings of the International Conference on Intelligent Data Engi-
neering and Automated Learning. Springer, 262–270.
[179] Eva Tuba, Raka Jovanovic, Romana Capor Hrosik, Adis Alihodzic, and Milan Tuba. 2018. Web intelligence data clus-
tering by bare bone fireworks algorithm combined with K-means. In Proceedings of the 8th International Conference
on Web Intelligence, Mining and Semantics. ACM, 7.
[180] Eva Tuba, Ivana Strumberger, Nebojsa Bacanin, and Milan Tuba. 2018. Bare bones fireworks algorithm for capaci-
tated p-median problem. In Proceedings of the International Conference on Swarm Intelligence. Springer, 283–291.
[181] Eva Tuba, Ivana Strumberger, Dejan Zivkovic, Nebojsa Bacanin, and Milan Tuba. 2018. Rigid image registration by
bare bones fireworks algorithm. In Proceedings of the 2018 6th International Conference on Multimedia Computing
and Systems (ICMCS’18). IEEE, 1–6.
[182] Eva Tuba, Ira Tuba, Diana Dolicanin-Djekic, Adis Alihodzic, and Milan Tuba. 2018. Efficient drone placement for
wireless sensor networks coverage by bare bones fireworks algorithm. In Proceedings of the 6th International Sym-
posium on Digital Forensic and Security (ISDFS’18). IEEE, 1–5.

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
121:26 J. Li and Y. Tan

[183] Eva Tuba, Milan Tuba, and Marko Beko. 2016. Node localization in ad hoc wireless sensor networks using fireworks
algorithm. In Proceedings of the 2016 5th International Conference on Multimedia Computing and Systems (ICMCS’16).
IEEE, 223–229.
[184] Eva Tuba, Milan Tuba, and Marko Beko. 2016. Support vector machine parameters optimization by enhanced fire-
works algorithm. In Proceedings of the International Conference in Swarm Intelligence. Springer, 526–534.
[185] Eva Tuba, Milan Tuba, and Edin Dolicanin. 2017. Adjusted fireworks algorithm applied to retinal image registration.
Stud. Inf. Contr. 26, 1 (2017), 33–42.
[186] Eva Tuba, Milan Tuba, and Dana Simian. 2016. Wireless sensor network coverage problem using modified fire-
works algorithm. In Proceedings of the 2016 International Wireless Communications and Mobile Computing Conference
(IWCMC’16). IEEE, 696–701.
[187] Eva Tuba, Milan Tuba, Dana Simian, and Raka Jovanovic. 2017. JPEG quantization table optimization by guided
fireworks algorithm. In Proceedings of the International Workshop on Combinatorial Image Analysis. Springer, 294–
307.
[188] Milan Tuba, Nebojsa Bacanin, and Adis Alihodzic. 2015. Multilevel image thresholding by fireworks algorithm. In
Proceedings of the 2015 25th International Conference Radioelektronika (RADIOELEKTRONIKA’15). IEEE, 326–330.
[189] Milan Tuba, Nebojsa Bacanin, and Marko Beko. 2015. Fireworks algorithm for RFID network planning problem. In
Proceedings of the 2015 25th International Conference Radioelektronika (RADIOELEKTRONIKA’15). IEEE, 440–444.
[190] Viktor Tuba, Adis Alihodzic, and Milan Tuba. 2017. Multi-objective RFID network planning with probabilistic cov-
erage model by guided fireworks algorithm. In Proceedings of the 2017 10th International Symposium on Advanced
Topics in Electrical Engineering (ATEE’17). IEEE, 882–887.
[191] Chunzhi Wang, Pan Wu, Lingyu Yan, Fangyu Zhou, and Wencheng Cai. 2018. Image retrieval based on fireworks
algorithm optimizing convolutional neural network. In Proceedings of the 2018 IEEE 4th International Symposium
on Wireless Systems within the International Conferences on Intelligent Data Acquisition and Advanced Computing
Systems (IDAACS-SWS’18). IEEE, 53–56.
[192] Xiaojing Wang, Hu Peng, Changshou Deng, Lixian Li, and Likun Zheng. 2018. An improved firefly algorithm hybrid
with fireworks. In Proceedings of the International Symposium on Intelligence Computation and Applications. Springer,
27–37.
[193] Yue Wang, Min-Xia Zhang, and Yu-Jun Zheng. 2017. A hyper-heuristic method for UAV search planning. In Pro-
ceedings of the International Conference in Swarm Intelligence. Springer, 454–464.
[194] Zhenyu Wang, Qibing Zhu, Min Huang, and Bao Yang. 2017. Optimization of economic/environmental operation
management for microgrids by using hybrid fireworks algorithm. Int. Trans. Electr. Ener. Syst. 27, 12 (2017).
[195] Xing Wei, Lei Liu, Yongji Wang, and Ye Yang. 2018. Reentry trajectory optimization for a hypersonic vehicle based
on an improved adaptive fireworks algorithm. Int. J. Aerosp. Eng. (2018).
[196] Zhenchun Wei, Liangliang Wang, Zengwei Lyu, Lei Shi, Meng Li, and Xing Wei. 2018. A multi-objective algorithm
for joint energy replenishment and data collection in wireless rechargeable sensor networks. In Proceedings of the
International Conference on Wireless Algorithms, Systems, and Applications. Springer, 497–508.
[197] Chengkai Xia, Zhenchun Wei, Zengwei Lyu, Liangliang Wang, Fei Liu, and Lin Feng. 2018. A novel mixed-variable
fireworks optimization algorithm for path and time sequence optimization in WRSNs. In Proceedings of the Interna-
tional Conference on Communicatins and Networking in China. Springer, 24–34.
[198] Zeyun Xiao, Bin Tian, and Xiaochun Lu. 2019. Locating the critical slip surface in a slope stability analysis by
enhanced fireworks algorithm. Clust. Comput. 22, 1 (2019), 719–729.
[199] Suchao Xie, Haihong Li, Chengxing Yang, and Shuguang Yao. 2018. Crashworthiness optimisation of a compos-
ite energy-absorbing structure for subway vehicles based on hybrid particle swarm optimisation. Structural and
Multidisciplinary Optimization 58, 5 (2018), 2291–2308.
[200] Bo Xing and Wen-Jing Gao. 2016. Innovative Computational Intelligence: A Rough Guide to 134 Clever Algorithms.
Springer.
[201] Junjie Xue, Ying Wang, and Jiyang Xiao. 2017. Uncertain bilevel knapsack problem and its solution. J. Syst. Eng.
Electr. 28, 4 (2017), 717–724.
[202] Jun-jie Xue, Ying Wang, Hao Li, Xiang-fei Meng, and Ji-yang Xiao. 2016. Advanced fireworks algorithm and its
application research in PID parameters tuning. Math. Probl. Eng. (2016).
[203] Jun-Jie Xue, Ying Wang, Hao Li, and Ji-yang Xiao. 2016. Discrete fireworks algorithm for aircraft mission planning.
In Proceedings of the International Conference in Swarm Intelligence. Springer, 544–551.
[204] Yu Xue, Binping Zhao, and Tinghuai Ma. 2016. Classification based on fireworks algorithm. In Bio-Inspired
Computing-Theories and Applications. Springer, 35–40.
[205] Yu Xue, Binping Zhao, Tinghuai Ma, and Alex X Liu. 2018. An evolutionary classification method based on fireworks
algorithm. Int. J. Bio-Insp. Comput. 11, 3 (2018), 149–158.

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
A Comprehensive Review of the Fireworks Algorithm 121:27

[206] Yu Xue, Binping Zhao, Tinghuai Ma, and Wei Pang. 2018. A self-adaptive fireworks algorithm for classification
problems. IEEE Access 6 (2018), 44406–44416.
[207] Weibo Yang and Liangjun Ke. 2019. An improved fireworks algorithm for the capacitated vehicle routing problem.
Front. Comput. Sci. 13, 3 (2019), 552–564.
[208] Wanchun Yang, Chenxi Zhang, and Bin Mu. 2015. Data-intensive service mashup based on game theory and hybrid
fireworks optimization algorithm in the cloud. Informatica 39, 4 (2015), 421–429.
[209] Xiang Yang and Ying Tan. 2014. Sample index based encoding for clustering using evolutionary computation. In
Proceedings of the International Conference in Swarm Intelligence. Springer, 489–498.
[210] Sengang Ye, Haiping Ma, Sheng Xu, Wenqiang Yang, and Minrui Fei. 2017. An effective fireworks algorithm for
warehouse-scheduling problem. Trans. Inst. Meas. Contr. 39, 1 (2017), 75–85.
[211] Wenwen Ye and Jiechang Wen. 2017. Adaptive fireworks algorithm based on simulated annealing. In Proceedings of
the 2017 13th International Conference on Computational Intelligence and Security (CIS’17). IEEE, 371–375.
[212] Xuemei Ye, Junzhi Li, Bo Xu, and Ying Tan. 2018. Which mapping rule in the fireworks algorithm is better for large
scale optimization. In Proceedings of the 2018 IEEE Congress on Evolutionary Computation (CEC’18). IEEE, 1–8.
[213] Xiaomeng Yin, Xinmin Li, Lei Liu, and YongJi Wang. 2017. Improved fireworks algorithm and its application in PID
parameters tuning. In Proceedings of the 2017 36th Chinese Control Conference (CCC’17). IEEE, 9841–9846.
[214] Xiaomeng Yin, Xing Wei, Lei Liu, and Yongji Wang. 2018. Improved hybrid fireworks algorithm-based parameter
optimization in high-order sliding mode control of hypersonic vehicles. Complexity (2018).
[215] Chao Yu, Lingchen Kelley, Shaoqiu Zheng, and Ying Tan. 2014. Fireworks algorithm with differential mutation for
solving the CEC 2014 competition problems. In Proceedings of the 2014 IEEE Congress on Evolutionary Computation
(CEC’14). IEEE, 3238–3245.
[216] Chao Yu, Ling Chen Kelley, and Ying Tan. 2015. Dynamic search fireworks algorithm with covariance mutation for
solving the CEC 2015 learning based competition problems. In Proceedings of the 2015 IEEE Congress on Evolutionary
Computation (CEC’15). IEEE, 1106–1112.
[217] Chao Yu, Ling Chen Kelley, and Ying Tan. 2016. Cooperative framework fireworks algorithm with covariance mu-
tation. In Proceedings of the 2016 IEEE Congress on Evolutionary Computation (CEC’16). IEEE, 1196–1203.
[218] Chao Yu, Junzhi Li, and Ying Tan. 2014. Improve enhanced fireworks algorithm with differential mutation. In Pro-
ceedings of the 2014 IEEE International Conference on Systems, Man and Cybernetics (SMC’14). IEEE, 264–269.
[219] Chao Yu and Ying Tan. 2015. Fireworks algorithm with covariance mutation. In Proceedings of the 2015 IEEE Congress
on Evolutionary Computation (CEC’15). IEEE, 1250–1256.
[220] Jun Yu and Hideyuki Takagi. 2017. Acceleration for fireworks algorithm based on amplitude reduction strategy and
local optima-based selection strategy. In Proceedings of the International Conference in Swarm Intelligence. Springer,
477–484.
[221] Jun Yu, Hideyuki Takagi, and Ying Tan. 2018. Accelerating the fireworks algorithm with an estimated convergence
point. In Proceedings of the International Conference on Swarm Intelligence. Springer, 263–272.
[222] J. Yu, H. Takagi, and Y. Tan. 2018. Multi-layer explosion based fireworks algorithm. J. Swarm Intel. Evol. Comput. 7,
173 (2018), 2.
[223] Jun Yu, Ying Tan, and Hideyuki Takagi. 2018. Scouting strategy for biasing fireworks algorithm search to promising
directions. In Proceedings of the Genetic and Evolutionary Computation Conference Companion. ACM, 99–100.
[224] Marcin Zalasiński, Krystian Łapa, and Krzysztof Cpałka. 2018. Prediction of values of the dynamic signature features.
Exp. Syst. Appl. 104 (2018), 86–96.
[225] Bei Zhang, Minxia Zhang, and Yu-Jun Zheng. 2014. Improving enhanced fireworks algorithm with new gaussian
explosion and population selection strategies. In Proceedings of the International Conference in Swarm Intelligence.
Springer, 53–63.
[226] Bei Zhang, Min-Xia Zhang, and Yu-Jun Zheng. 2014. A hybrid biogeography-based optimization and fireworks
algorithm. In Proceedings of the 2014 IEEE Congress on Evolutionary Computation (CEC’14). IEEE, 3200–3206.
[227] Bei Zhang, Yu-Jun Zheng, Min-Xia Zhang, and Sheng-Yong Chen. 2017. Fireworks algorithm with enhanced fire-
works interaction. IEEE/ACM Trans. Comput. Biol. Bioinform. 14, 1 (2017), 42–55.
[228] Jun Zhang and Hang Zhang. 2018. An improved back propagation neural network forecasting model using variation
fireworks algorithm for short-time traffic flow. In Proceedings of the 2018 13th World Congress on Intelligent Control
and Automation (WCICA’18). IEEE, 1085–1090.
[229] JunQi Zhang, ShanWen Zhu, and MengChu Zhou. 2017. From resampling to non-resampling: A fireworks algorithm-
based framework for solving noisy optimization problems. In Proceedings of the International Conference in Swarm
Intelligence. Springer, 485–492.
[230] Min Zhang, Yi Yuan, Ruiqi Wang, and Wenming Cheng. 2018. Recognition of mixture control chart patterns based
on fusion feature reduction and fireworks algorithm-optimized MSVM. Pattern Anal. Appl. (2018), 1–12.

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.
121:28 J. Li and Y. Tan

[231] Qingfu Zhang and Hui Li. 2007. MOEA/D: A multiobjective evolutionary algorithm based on decomposition. IEEE
Trans. Evol. Comput. 11, 6 (2007), 712–731.
[232] Tianjiao Zhang, Liangjun Ke, Jisheng Li, Jing Li, Zexi Li, and Jingqi Huang. 2016. Fireworks algorithm for the satellite
link scheduling problem in the navigation constellation. In Proceedings of the 2016 IEEE Congress on Evolutionary
Computation (CEC’16). IEEE, 4029–4037.
[233] Tingting Zhang and Zhifeng Liu. 2017. Fireworks algorithm for mean-VaR/CVaR models. Physica A 483 (2017), 1–8.
[234] Yuqiong Zhang, Jiale Liu, Huizhi Zhou, Ke Guo, and Fei Tang. 2018. Intelligent reconfiguration for distributed power
network with multivariable renewable generation. In Proceedings of the 2018 Asian Conference on Energy, Power and
Transportation Electrification (ACEPT’18). IEEE, 1–5.
[235] Xinchao Zhao, Rui Li, Xingquan Zuo, and Ying Tan. 2017. Elite-leading fireworks algorithm. In Proceedings of the
International Conference in Swarm Intelligence. Springer, 493–500.
[236] Shaoqiu Zheng, Andreas Janecek, Junzhi Li, and Ying Tan. 2014. Dynamic search in fireworks algorithm. In Proceed-
ings of the 2014 IEEE Congress on Evolutionary Computation (CEC’14). IEEE, 3222–3229.
[237] Shaoqiu Zheng, Andreas Janecek, and Ying Tan. 2013. Enhanced fireworks algorithm. In Proceedings of the 2013 IEEE
Congress on Evolutionary Computation (CEC’13). IEEE, 2069–2077.
[238] Shaoqiu Zheng, Junzhi Li, Andreas Janecek, and Ying Tan. 2017. A cooperative framework for fireworks algorithm.
IEEE/ACM Trans. Comput. Biol. Bioinf. 14, 1 (2017), 27–41.
[239] Shaoqiu Zheng, Lang Liu, Chao Yu, Junzhi Li, and Ying Tan. 2014. Fireworks algorithm and its variants for solving
ICSI2014 competition problems. In Proceedings of the International Conference in Swarm Intelligence. Springer, 442–
451.
[240] Shaoqiu Zheng and Ying Tan. 2013. A unified distance measure scheme for orientation coding in identification. In
Proceedings of the 2013 International Conference on Information Science and Technology (ICIST’13). IEEE, 979–985.
[241] Shaoqiu Zheng, Chao Yu, Junzhi Li, and Ying Tan. 2015. Exponentially decreased dimension number strategy based
dynamic search fireworks algorithm for solving CEC2015 competition problems. In Proceedings of the 2015 IEEE
Congress on Evolutionary Computation (CEC’15). IEEE, 1083–1090.
[242] Yu-Jun Zheng, Qin Song, and Sheng-Yong Chen. 2013. Multiobjective fireworks optimization for variable-rate fer-
tilization in oil crop production. Appl. Soft Comput. 13, 11 (2013), 4253–4263.
[243] Yu-Jun Zheng, Xin-Li Xu, Hai-Feng Ling, and Sheng-Yong Chen. 2015. A hybrid fireworks optimization method with
differential evolution operators. Neurocomputing 148 (2015), 75–82.

Received November 2018; revised September 2019; accepted September 2019

ACM Computing Surveys, Vol. 52, No. 6, Article 121. Publication date: November 2019.