Accepted Manuscript

Socio evolution & learning optimization algorithm: A socio-inspired

optimization methodology

Meeta Kumar, Anand J. Kulkarni, Suresh Chandra Satapathy

PII: S0167-739X(17)31725-9
DOI: https://doi.org/10.1016/j.future.2017.10.052
Reference: FUTURE 3791

To appear in: Future Generation Computer Systems

Received date : 31 July 2017

Revised date : 29 September 2017
Accepted date : 29 October 2017

Please cite this article as: M. Kumar, A.J. Kulkarni, S.C. Satapathy, Socio evolution & learning
optimization algorithm: A socio-inspired optimization methodology, Future Generation Computer
Systems (2017), https://doi.org/10.1016/j.future.2017.10.052

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form.
Please note that during the production process errors may be discovered which could affect the
content, and all legal disclaimers that apply to the journal pertain.
 Socio Evolution & Learning Optimization Algorithm: A
Socio-inspired Optimization Methodology
Meeta Kumar1, Anand J. Kulkarni1, 2*, Suresh Chandra Satapathy3
1
Symbiosis Institute of Technology, Symbiosis International University, Pune, MH 412 115, India
Email: {meeta.kumar; anand.kulkarni}@sitpune.edu.in
2
Odette School of Business, University of Windsor, 401 Sunset Avenue, Windsor, ON N9B3P4,
Canada
Email: [email protected]; Ph: +1 519 253 3000 (Ext: 4939)
3
Department of Computer Science and Engineering, PVP Siddhartha Institute of Technology,
Vijayawada, AP, India
Email: [email protected]

A bstract: The paper proposes a novel metaheuristic Socio Evolution & Learning Optimization Algorithm (SELO)
inspired by the social learning behaviour of humans organized as families in a societal setup. This population based
stochastic methodology can be categorized under the very recent and upcoming class of optimization algorithms- the
socio-inspired algorithms. It is the social tendency of humans to adapt to mannerisms and behaviors of other individuals
through observation. SELO mimics the socio-evolution and learning of parents and children constituting a family.
Individuals organized as family groups (parents and children) interact with one another and other distinct families to
attain some individual goals. In the process, these family individuals learn from one another as well as from individuals
from other families in the society. This helps them to evolve, improve their intelligence and collectively achieve shared
goals. The proposed optimization algorithm models this de-centralized learning which may result in the overall
  


  
 




 

re societal system. SELO shows
good performance on finding the global optimum solution for the unconstrained optimization problems. The problem
solving success of SELO is evaluated using 
well-known boundary-constrained benchmark test problems. The paper
compares the results of SELO with few other population based evolutionary algorithms which are popular across
scientific and real-world applications. 
  




nother very recent socio-inspired
methodology-the Ideology algorithm. Results indicate that SELO demonstrates comparable performance to other
comparison algorithms. This gives ground to the authors to further establish the effectiveness of this metaheuristic by
solving purposeful and real world problems.

K eywords: cultural evolution; evolutionary algorithm; metaheuristic; nature-inspired computing; socio-inspired

optimization; unconstrained optimization.


1 Introduction and Motivation

Optimization methods play a vital role in solving engineering problems. The exact optimization methods or
deterministic methods may not be computationally efficient in solving complex nonlinear and multimodal
problems that exist in most real-world applications (Luke, 2013; Talbi, 2009). In the past few decades
researchers have resorted to a number of methodologies inspired from biological and natural systems have been
proposed for solving complex optimization problems. By far the majority of nature-inspired algorithms (Yang,
2010) are based on certain characteristics of biological system (Brownlee, 2011; Fister, 2013). Recent literature
(Nanda and Panda, 2014; Zang et al., 2010) advocate that the largest number of nature-inspired algorithms is
categorized as bio-inspired algorithms. Closer to the bio-inspired algorithms are the swarm based algorithms
(Bonabeau et al., 1999; Hassanien and Emary, 2016) which seek inspiration from the collective or swarm
behavior exhibited by a number of animal species. The collective intelligence as exhibited by a swarm towards
achievement of shared goal forms the key idea behind swarm-based algorithms. A large number of
homogeneous agents in an environment function through mutual cooperation to achieve desired goal. Another
popular category is the Evolutionary Algorithms (EA) which may be categorized under bio-inspired algorithms
and seek inspiration from the processes of biological evolution and natural selection. Few examples include
genetic algorithms (Bonabeau et al., 1999), ant colony optimization (Dorigo et al., M., 2006) based on the
interaction of ants, fish school search (Filho et al. 2009) based on the collective behavior of fishes in order to
survive and many others. Many algorithms find motivation from physical and chemical systems too, as reviewed
by Fister et al. (2013) and Biswas et al. (2013). A few examples include, but not limited to Simulated Annealing
(Brooks and Morgan, 1995; Goffe et al., 1994; Kirkpatrick et al., 1983) and Harmony Search (Geem, 2010;
Yang, 2009) which find their base in physical processes; while Chemical Reaction Optimization (Lam and Li
2010) is an example of chemical-inspired methodology. A very upcoming and emerging branch of EA is the
Socio-Inspired class of algorithms, which take motivation from the social and cultural interactions seen in
human behavior. A detailed review of this class of algorithms is presented in the further section. Thus the
existing nature-inspired algorithms may be briefly categorized into following broad or major categories: swarm
intelligence based, bio-inspired, physical /chemical inspired, and others (like social, cultural algorithms). Figure
 details the broad classification of these algorithms. The   further attempt to sub-classify the Social
Algorithms in figure  based on the broad ideas from which they draw inspiration. A list of Socio Inspired
algorithms is also presented in table
. All these methods have gained popularity because of their use of simple
rules for searching for optimal solutions to complex and real-world computational problems. Metaheuristics
(Gendreau and Potvin, 2010) is a term used to refer to such general algorithmic framework which seek
inspiration from various phenomenon observed in nature. Metaheuristic strategies combine rules and a certain
degree of randomness to find optimal (or near optimal) solutions to problems and thus can be applied to a
variety of optimization problems. They are general approximate strategies (Talbi, 2009) and can be adapted to
solve a wide variety of optimization problems with little changes to their general algorithmic framework. The
popularity of metaheuristics in solving real world optimization problems can be attributed to their following
characteristics: simple design owing to their closeness to natural concepts, less problem specific, faster problem
solving ability and ability to scale large dimension problems. The development of these metaheuristics has been
a thrust area in the field of computational intelligence for decades now, with newer class of algorithms fast
evolving, with well-accepted algorithms like Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Ant
Colony Algorithms and socio algorithms like Cohort Intelligence among others.
Originally proposed by Kennedy and Eberhart in 1995, PSO is based on the behavior exhibited by flocking birds
or a school of fish. In PSO (Marini and Walczak, 2015), swarm of birds/fish representing the candidate solutions
travel through the sample space driven by their own and best performances of their neighbors. Although
popular, PSO suffers from getting stuck in local minima and premature convergence for certain complex
problems. The comprehensive learning PSO (CLPSO) (Liang et al., 2006) and PSO2011 (Clerc and Maurice
2012; Omran and Clerc, 2011) are the improved variants of the standard PSO partially overcoming this
limitation. CLPSO maintains the diversity of the swarm by allowing a particle to update its velocity by using
previous best positions of other particles. Each dimension of a particle seeks learning from a different particle.
The EA draw inspiration from 
 


 
analogy as stated in the theory of natural selection by
Charles Darwin. 
  
 very popular GA and the others such as Evolution Strategies, Genetic
Programming, Evolutionary Programming and Differential Evolution; as elaborate in the article by Nanda
(2014). Covariance matrix adaptation evolution strategy (CMAES) (Hansen, 2006; Igel et al., 2007) is a
powerful EA for real-valued optimization based on self-adaptation in evolution strategies. The artificial bee
colony algorithm (ABC) (Karaboga and Basturk, 2007 and 2008) is an optimizer based on the collaborative and
intelligent behavior of a swarm of honey bees. It mimics the foraging behavior of swarm of bees which
comprises of employed bees, onlooker and scout bees. It is a powerful metaheuristic which effectively employs
four different selection processes: a global selection process , local selection process carried out by employed
and onlooker bees, a local greedy selection process to memorize the best candidate solution and a random
selection process carried out by the scouts. Differential evolution (DE) algorithm, proposed by Storn and Price




(1997) is a powerful yet simple population-based optimization technique mimicking the basic rules of genetics,
yet different from the basic Genetic algorithm. It employs (Neri and Tirronen, 2010) five mutation and two
crossover strategies. The Adaptive DE algorithm (JDE) (Zhang and Sanderson, 2009) and the Self-Adaptive
Differential Evolution algorithm (SADE) (Brest, 2006; Qin et al., 2005) are modified versions of the DE. The
JDE provides a modified DE algorithm with self-tuned control parameters. In SADE, both, the control
parameters as well as the learning strategies of DE are self-adapted during the evolution phase. Another
optimization methodology is the Backtracking Search Algorithm (BSA) (Civicioglu, 2013) exploits certain
genetic operators such as selection, mutation and crossover. However it uses a random mutation approach where
an individual from the previous population is retained, which helps it to seek important learning from the
experiences of the previous population. Also, the crossover strategy of BSA is more complex than the crossover
strategies used across the various derivative forms of GA. Occasionally new metaheuristics are introduced
which use a novel metaphor as guide for solving optimization problems. Socio-inspired algorithms are one such
very recent and upcoming class of optimization algorithms, which use the idea of simulating and mimicking
social learning of humans (or social evolution). The notion of Cultural Algorithms (Reynolds and Sverdlik,
1994) was first introduced by Reynolds as early as  which states that individuals evolve much faster
through cultural evolution than through biological or genetic evolution alone. Humans adapt to mannerisms and
behaviors by observing/imitating other individuals which helps them improve their intelligence quickly and
achieve shared goals. The tendency to cooperate and function together as a cohesive group adds to their
collective intelligence. This idea has formed the ground for many a researchers to formalize some recent socio-
inspired algorithms like Society and Civilization Optimization algorithm(SCO), Imperialist Competitive
algorithm (ICA), League Championship algorithm (LCA), Social Emotional Optimization algorithm (SEOA),
Election Campaign Optimization algorithm (ECO), Anarchic Society Optimization algorithm (ASO), Teaching

learning-based optimization (TLBO), Cultural Evolution algorithm (CEA), Cohort Intelligence (CI), Soccer
League Competition Algorithm (SLC), Election Algorithm, Social learning optimization (SLO) , Social Group
Optimization (SGO) and Ideology Algorithm (IA) etc.
SCO (Ray and Liew, 2003) is inspired from human social behavior seen among society individuals. The
individuals in a society interact with one another to improve their overall behavior and a cooperative interaction
among such societies represents a civilization. The ICA (Atashpaz-Gargari and Lucas, 2007; Hosseini and
Khaled, 2014) was originally proposed in 2007, simulates the socio-political behaviors seen across imperialist
nations which compete to take possession of weaker colonies or empires. This imperialist competition finally
results in strengthening the power of stronger and successful imperialist empires; whilst the weaker empires
collapse gradually finally leading to a state of convergence. LCA established by Kashan (2009, 2011 and 2014)
is derived from the competition amongst teams seen in league matches. Artificial teams (representing solutions)
compete across weeks (representing iterations) based on a league schedule and a strong team (with higher
fitness value) gradually emerges as the winner at the end of the playing season (stopping condition). SEOA (Xu
et al., 2010) is a swarm based socio-inspired metaheuristic which simulates a virtual individual person who
wishes to achieve a higher status in the society and his decisions are guided by his emotion. An emotional index
determines his current behavior and the society determines whether his current behavior is better or worse, thus
affecting this emotion index value (parameter). The authors Lv et al. (2010) introduced the ECO algorithm
based on the mannerisms of political candidates during an election campaign. The voters are inspired to vote for
a candidate with better prestige (better function value) and finally a stronger election candidate wins the highest
supports from voters (Xie et al., 2010). Another optimization algorithm which seeks inspiration from a very
commonly observed human behavior of being greedy and disorderly to achieve their goals is ASO algorithm
(Ahmadi-Javid, 2011). Members behave anarchically to find better solutions in the solution space. TLBO (Rao
et al., 2012) an optimization method mimics the influence of a teacher on the learning outcome of it students.
The methodology works in two parts: the teacher phase and the learner phase, where the learners increase their
knowledge (improve solution quality) by imbibing knowledge from the teacher as well through interaction with
their peer learners. Kuo and Lin presented a framework for Cultural Evolution Algorithm in
. They state
that a species learns or evolves through one of these modes: group consensus, individual learning, innovative
learning and self-improvement and in their research mathematically model each of these learning modes. CI
(Kulkarni et al., 2016; Kulkarni and Shabir, 2016) a successful socio-inspired metaheuristic mimics the self-
learning behavior exhibited by candidates in a group, where the candidates cooperate and compete with one
another to achieve some individual goal. SLC (Moosavian, 2015 and 2016) has been effectively applied to solve
discrete and continuous optimization problems and takes inspiration from competitive behavior seen among
teams and players in soccer league matches. Election Algorithm (Emami and Derakhshan, 2015), yet another
optimization and search algorithm, tries to mimic the political elections comprising the electoral parties, the
candidates and the voters. SLO (Liu et al., 2016) which proposes an optimization methodology where
individuals evolve at different levels: one is genetic evolution and the other is cultural evolution through
imitation and observation. This cultural evolution also in turn influences the genetic evolution in future
generations to come and helps accelerate human intelligence. A very recent algorithm, the Social Group





Optimization (Satapathy and Naik, 2016) takes motivation from the human behavior exhibited when trying to
collectively attempt to solve a complex task at hand. This mannerism of getting influenced by a better person;
modifying his behavior accordingly helps address complex problems. Another innovative algorithm IA (Huan et
al., 2016) is inspired from the idea how certain beliefs become the guide for individuals in a society to achieve
their goals. IA elicits this idea through a political scenario where individuals follow their political ideologies and
compete with members of their political party as well as with leaders of other political parties in their will to
excel. It can thus be seen from the literature that a number of popular and efficient optimization algorithms have
been proposed and will continue to develop; however in agreement with no free lunch in optimization it is true
that one particular optimization algorithm may be more suited to some optimization problem better than others.
It underscores that a universal algorithm/method does not exist for all optimization problems and that on an
average will perform equally well. Typically for a defined problem with specific objective functions there will
be a class of optimization algorithms that will outperform some others. The major task at hand is only to identify
these better performing algorithms which are specialized to the structure of specific optimization problem under
consideration. Therefore, it can be stated that developing new optimization algorithms will be always essential
and significant and gives the authors a ground that there will be always a scope to develop newer prospective
algorithms that could be well-suited to some specific class of optimization tasks and may as well surpass a few
other already existing algorithms for solving some specific optimization problems (Ho and Pepyne, 2002).
In the current research, the authors propose a novel socio-inspired optimization methodology referred to as
Social Evolution & Learning Optimization algorithm. The metaheuristic is motivated and draws inspiration
from the social behavior exhibited by individuals in a family which is a part of a human societal setup. A family
represents an elementary social group in a society typically consisting of parents and their children and a society
can be visualized as a multi-agent setup of different families coexisting together. According to Goldsmith
(1978)  










  


  Each family
member can be thought of as an individual agent in a family, making its own behavioral choices inspired by
observing and learning from others. The evolution of parents and children of a family is based on learning from
one another as well as from other families. This de-centralized learning may result in the overall improvement


 
 




 


 
 
The proposed
optimization algorithm models the above rationale of   
 
 




individuals who collectively evolve their family. Thus it may be stated that SELO takes inspiration and mimics
natural and social tendency of humans organized as family groups, where individual family members (parents
and kids) interact with one another and other distinct families to achieve some individual/shared goals. A kid is
genetically similar to its parent and exhibits similar behaviour (fitness); however may later be influenced by the
social behaviour of his peers. Unlike traditional  the proposed methodology provides for a two-way
evolution; allowing for a two way system of learning. In SELO, evolution and learning takes place at both:
social level and population level (for every individual).
In a multi-agent environment (Kulkarni and Tai, 2010), the algorithm tries to simulate the behaviour of agents
(family members) who work in a coordinated way, optimizing their local utilities and contributing the maximum
towards optimization of the global objective (betterment of the society as a whole).According to Hechter and
Horne (2009)  

 
 


 




SELO follows this
very notion and uses intra and intersociety interactions observed within human society to converge to an optimal
solution of the optimization problem at hand. In the context of human society, one family unit may be inspired
or motivated to follow and pursue the qualities of the individuals of the other families, which in-turn, may result
in the improvement of its own qualities and the associated behaviour. The parents from a family, living in a
societal setup may be influenced by the positive/negative qualities and behaviour of the parents from other
families. Also, all the individuals of a family typically may exhibit similar behaviour and have similar choices
and likings. Qualities of the parents which construct their social behaviours is typically assimilated or followed
by their children also. So during the process of socialization, every member of the family tries to advance its
own behaviour by either observing its own family or other families. When a family member attempts to follow a
given behaviour characterized by certain qualities, it often adopts such qualities in a manner that may improve
its own behaviour and associated goals. This socio-behavioural model using SELO enabled to solve a multitude
of optimization test problems with diverse properties so that the usefulness of the proposed method can be truly
tested in an unbiased way. The authors compare the problem solving success and robustness of the proposed
methodology with some popular and well-established metaheuristics (Civicioglu and Besdok, 2013) like PSO
(its variants PS O2011 and CLPS O ), CMAES, ABC and BSA as well with newer and promising socio-inspired
methodology like IA. This gives a clear picture of how the proposed algorithms SELO fairs at solving numerical
optimization problems in comparison to varied classes of other evolutionary algorithms.
The remainder of the paper is organized as follows. Section
discusses the exact methodology and algorithmic
framework for the proposed technique and how the idea can be applied to solving optimization problems.





Section  tabulates the computational results obtained and discusses the performance of SELO and other
comparator algorithms. Finally, section  concludes the paper and suggests future directions for researchers.

Nature Inspired Algorithms

Bio-Inspired Algo. Swarm Intelligence Physical ,

Algo. Chemical Systems
Based Algo.

Artifical Immune Ant Colony

Evolutionary Algo. System Cat Swarm Opt.
Simulated
Bacteria Foraging Cuckoo Search
Annealing ,
& many others Firefly Algo
Harmony Search
Bat algorithm

Genetic Algorithm
Evolution Strategies
C ultural / Social
Genetic Programming A lgorithm ( SA)
Evolutionary Programming
Differential Evolution

F ig. 1. Broad Classification of Nature-Inspired Algorithms

In this way, parents and kid from a family may learn from one another which may help the entire society evolve
and improve.
2 M ethodology : Socio E volution & L earning O ptimization A lgorithm (SE L O)
The proposed iterative algorithm is population based, which initially starts its search and optimization process
with a population of solutions. Akin to other population-based designs, SELO attempts to direct the population
of possible solutions towards the more promising areas of the solution space in search for optimal solution. In
the context of SELO, the behaviour of an individual belonging to a family represents each such solution. Each
family comprises of individuals or family members, who can either, be a parent or a kid. Thus members of the
immediate family include two parents, and sons and/or daughters (referred to as kids in the algorithm).Thus
population comprises a set of multiple families.
This social framework comprising families and a cluster of families (a society) forms the basis of our proposed
socio-inspired SELO. In the context of an optimization problem, the objective function represents the behaviour
  of a family individual defined as
  


  
 

, and the variables 


  
 


represent the qualities of each family individual. We present certain characteristics observed in a societal setup
of human families to describe the artificial Socio Evolution & Learning Algorithm which will be explained in
the further sections of the paper.

An individual represents a basic element of a society.

A certain group of individuals coexist as a family.

A society can be thought of as a group of individuals or group of families involved in continual social
interaction.

Every family in the population (considering a typical setup) comprises of individuals or family
members, who are either a parent or a kid. Thus members of the immediate family may include two
parents (grownups), and their offspring (sons and/or daughters).

The grownups or elders follow behaviours and get influenced (evolve) by the mannerisms of other
individuals, during the course of social interaction with other families.

Families provide initial socialization for children that shape their attitudes, values and behaviours.






A kid is genetically similar to its parents and will imbibe their behaviour in the earlier years and
exhibits similar fitness; however at behaviour level may later evolve differently inspired by social
behaviour of others (peers).

In the later or their growing years, children are greatly influenced by their peers and assimilate a lot of
peer behaviour in social settings in order to be accepted by their peers (Eisenberg, 2008).
Thus every human (an elder or a child) constantly learns through observation and imitation. This idea is strongly
supported through the social learning theory proposed by Albert Bandura (1962 and 1977) that human behaviour
is greatly due to genetic as well as physiological factors and is achieved primarily through learning. The
acquired experience environment and cultural influence greatly impacts his behaviour. Various social researches
and studies have also found that parents influence at-home behaviour of children (Maccoby, 1992), and peers
influence behaviour outside home or their social behaviour. This social learning accumulated from his parents
and his peers guide an 
  genetic evolution. As characterized from above, the SELO algorithm will
begin with the creation of initial population in some way, e.g. through randomization, with some initial number
of families; every family consisting of two parents and kid. A set of parents from each of the initialized
families generates their set of behaviour randomly in close neighbourhood of each other. Initially, every kid
belonging to a family generates its behaviour function correspondingly in the neighbourhood of their parents.
The algorithm then progresses iteratively where members from each family may decide to follow the behaviour
of its own family or other families. In the current study, the choice to follow is decided through a roulette wheel
approach (Goldberg and Deb, 1991; Pencheva et al., 2009) thus giving a fair chance to every behaviour in the
population to get selected purely based on its quality. A roulette wheel selection (RWS) operator is used to
recommend a fitter or better behaviour to either a parent or a kid. The parent or the kid may then choose to
follow this behaviour or fitness (detailed in step  of the algorithm). RWS is a fitness proportionate selection
where the selection mechanism abandons none of the individuals in the population and every individual gets a
fair opportunity to get selected, also ensuring some level of diversity in the population of solutions. Each
individual is assigned a share in the roulette wheel which is proportional to its fitness value (Kumar, 2012).
Then probabilistically roulette wheel returns the selected individual, with more probability that fitter individuals
may be chosen as compared to individuals with poor fitness. This selection algorithm thus provides a zero bias.
It also helps the algorithm jump out of possible local minima.

C ultural / Social A lgorithm (SA)

SA based on Socio- SA based on Sports SA based on Social and SA based on

Political Ideologies Competitive Behavior Cultural Interaction Colonization

Soccer League Cohort Intelligence Society and civilization

Ideology Algorithm Competition Algorithm Optimization Algorithm

Teaching Learning Imperialist Competitive

Election Algorithm League Championship Optimization Algorithm
Algorithm
Social Group
Anarchic Society
Election Campaign Optimization
Optimization
Algorithm
Social Learning
Optimization

Cultural Evolution
Algorithm
Social Emotional
Optimization

Socio E volution &

L earning O ptimization

F ig. 2 : Broad Classification of Socio-Inspired Algorithms

The concept of SELO is represented in figure
. Certain fitter individuals (with better behaviour) are selected
based on a fitness function evaluation which guides other individuals to gradually improve their behaviour.
Knowledge of fitter individuals thus guides other individuals in the society across iterations through an




  

  
 
 
  
 




 
 


  



 

 
 
  


 

  

 
 



improvement in the behaviour of every individual member in the families. The basic steps and mathematical
formulation of the algorithm are presented in section
 .

Algorithm Author Year, Algorithm Author Year,

Reference Reference
Ideology Algorithm Huan et al. 2016, [29] Teaching Learning Based Rao et al. 2012, [62]
Optimization
Social Learning Liu et al. 2016, [49] Anarchic Society Optimization Ahmadi 2011, [1]
Optimization
Social Group Optimization Satapathy et al. 2016, [65] Social Emotional Optimization Xu et al. 2010, [71]
Algorithm
Election Algorithm Emami et al. 2015, [15] Election Campaign Optimization Lv et al. 2010, [51]
Algorithm
Soccer League Competition Moosavian et al. 2014, [55] League Championship Algorithm Kashan 2009, [34]
Algorithm
Cultural Evolution Kuo et al. 2013, [46] Imperialist Competitive Atashpaz et 2007, [2]
Algorithm Algorithm al.
Cohort Intelligence Kulkarni et al. 2013, [42] Society And Civilization Ray et al. 2003, [63]

Table 1 A list of Socio-Inspired Algorithms

2.1 Steps and flowchart of the SE L O A lgorithm

Consider a generalized unconstrained problem (in minimization sense) as follows:
Minimize KL @ K
#
   1
   '
L
Subject to /26+4
1
C 1 C
533+4
1

  @    
The SELO procedure begins with initialization of number of families  where every family ,
 K @
   L
consists of two parents , )8  , )9
and kids
, &>
K @
   L K @    ), i.e.
)8 )9 &8 &> &;
,
D,
 ,  ,    ,    , E. Every individual  D#  $  #    -    ( E of each family ",
 K @

   L generates variables / qualities,

%= ? %= ? %= ? "= ?

@
H#    1
 
' I


 @     which makes its behaviour
K L. The individual behaviour of
every individual  of each family ",
 K @
   L is generally being observed by every individual from the
same as well as other families. This urges every family individual to follow the behaviour better than its current
behaviour. More specifically,  may follow J

J
? ?
if
 O "= P B
H "= I  A J
 




J
 D#  $  #    -    ( E .

In addition, initialize convergence parameter !, sampling interval reduction factor  M  N,

parent_follow_probability 3 M  N to a significantly large value and peer_follow_probability . M  N to a
small value, learning iteration counter , learning attempt counter  @ 0*7  

 

 
  

till required learning behaviour converges. Tuning the parameters 3 M  N and . M  Nallows the choice of
whether the kids follow the behaviour of their parents or their peers. Values of parameters 3
 . will be reduced
across iterations which signify that the kids gradually tend to get influenced or follow the behaviour of their
compeers (i.e. their own siblings or kids from other families).
Step 1 (Initialize families and objective function evaluation)
Every parent , )8 and
, )9 associated with family , K @    
Lrandomly generates its set of
<8 <8 <8 <8 <9 <9 <9 <9
qualities/variables
 %= @
O# %=  1 %=
 
' %= P and  %= @
O# %=  1 %=
 
' %= P in the close
neighbourhood of each other, from within the corresponding sampling interval F/26+4
1
 533+4
1
G



 @    
<8 <9
and further evaluates associated objective function/behaviour  O %= P and  O %= P, respectively.  is the
dimension of the optimization problem being solved by proposed algorithm.
:>
Every kid , &> ( @    

 @    L randomly generates its set of qualities/variables  %= @
:> :> :>

# %=  1 %=
 
' %= in the close neighbourhood of one of its parents and evaluates associated objective





 
function/behaviour
 &  '. This behaviour is illustrated as a schematic in the figure
, where a single family
is represented comprising of two parents with their two kids. Each member of the family generates its set of
variables " 

 # which denotes their qualities. It should be noted that this generation of initial qualities for
 
  

  
 


 function in MATLAB which generates random numbers whose
elements are uniformly distributed in the interval (,).

F amily 2
F amily 1


 
 



  

F amily 3

Parent 1  

  

Parent 2  

  


 

parents

Kid  

  


 

 
 

 

  


 

 

  
 
 
 

 

siblings families

F ig. 3 Schematic Representation of SELO

Step 2 (Parent Follow Behaviour / Parent Influence function)

Every parent   and
  associated with family  " ! 
# decides to follow corresponding behaviour
 
  $ %  $ %
 " #of one of the parent from certain family 
" !
#
nd associated qualities
!
  
 $ %
 $ %  $ %

   . Thus every parent may follow certain behaviour and the behaviour to follow is




selected based on roulette wheel selection approach (Kumar, 2012) and the subscript  indicates that the
behaviour is selected at random by family member and not known in advance. Following certain behaviour
implies that the current sampling space /interval of every variable associated with the parent are updated to the
close neighbourhood of the individual it follows. A parent may attempt to improve its behaviour by either
following the behaviour of another parent through an influence function or through a self-contemplation
operator. The self-contemplation operator (described in step
) here refers to searching in the close
neighbourhood of its own current sampling space and solution; which may result in self - improvement in its
own behaviour. Thus each parent of every family updates the current sampling interval associated with its every
$ $*+ $*+ $*+
% *+
variable
'
, %   %
 
 % -   '    to the close neighbourhood of the best solution
$
% *+
( ) this particular parent chooses to follow.

Step 3 (Kid Follow Behaviour / Kid Influence function)

Every kid  & ( '   

 '    ) associated with every Family  ( '   
) may choose to
follow his siblings or peers or may learn from the behaviours of elders in the societal setup. These elders may be
his parents; or in some cases, a kid may be influenced by the qualities of other parents or grownups from some
other family. Children will often replicate what they see, and not always what their parents ask them to do. The
above behaviour exhibited by kids is modelled using the Kid follow / influence function as follows:

Every parent  ! and
 " of family  randomly generates its variables 

 in close neighbourhood
of each other, from within the corresponding sampling interval

F amily 1, Parent 1 F amily 1, Parent 2

% $! % $! % $" % $"

 

 

V ariable  (chosen randomly) V ariable  (chosen randomly)

% #! % #! % #" % #"!

 

 

F amily



F amily 1, K id 2

! "
Every kid  and
 of family  generates its variable values 

 in close neighbourhood of its parents

F ig. 4. Generation of variable values for a family

Members from a better behaving family have greater influence on   

 
  
 
  a
kid from the global best family may mimic his parents or his siblings which in turn may lead to improvement in
his own behaviour. This observation in a social setting has been mapped using the probabilistic parameter
 ,
where 

is the parent_follow_probability and a kid may choose to follow his parents with a probability
 .





Every kid 0 -> I C   
 
 C   J from the global best family generates a random number ! 1 
K L and probabilistically performs the following:
If ! 1H D !4
then the kid 0 -> decides to follow the corresponding behaviour of either of its parents, else it
may choose to follow behaviour of one of its own sibling and using a roulette wheel approach decides to follow
;KL ;KL ;KL ;KL ;KL
,

the corresponding behaviour   I ,= J and associated qualities
 ,= C
"% =  "3 ,=  ". ,= of
either of his siblings. Thus every kid may follow certain behaviour and the superscript indicates that the
behaviour is selected at random by family member and not known in advance, the selection being based on
roulette wheel selection approach.
Some other kids may be largely influenced by their peers and may choose to follow other kids with a
probability
* . The parameter, kid_follow_probability * is significantly smaller in initial iterations and
increases as iterations progress simulating that the peer influence on the children grows as they age. Every other
kid ( '9 I C   
 
C   J from other families randomly chooses to follow behaviour from within
one of the other families and generates a random number  )  K L and probabilistically performs the
following:
If ! 1 D !2
then kid may choose to imbibe the behaviour of any one of its peers using a roulette wheel
; ;
,= KL ,= KL
approach and decides to follow the corresponding behaviour   I J nd associated qualities
C
; ; ;
, KL
,= KL ,= KL ->

"% =  "3  ". of the chosen peer, else the kid 0 randomly chooses to follow the behaviour
of one of the randomly chosen parents or elder from a selected family in the society using a roulette wheel
approach.
Step 4 (Sampling Interval Updation)
Following or mimicking certain behaviour (in steps  and
) implies that the current sampling interval of every
variable associated with every parent or kid is updated to the close neighbourhood of the individual it follows. A
parent may decide to follow the behaviour of another grown up from another family based on roulette wheel
selection approach (influence function) and calculates the updated objective function (or the associated updated
behaviour). Thus every individual of each family updates the sampling interval associated with its every
variable (refer to Equation IJ
 
IJ )

Every parent 0 /6 and
0 /7 associated with family 0 I C   
J samples variable values from within the
< <
, KL , KL
updated sampling interval $3=
C    associated with every variable  3= C    as follows:
: :
<
, KL
<
, KL
Q5?= Q <
, KL
Q5?= Q , ,
#3 =

O"3 = AM N
"3 = @
M N
P
, where #3 = C EG#3 = GF B !
















































IJ
& &

A parent may seek to improve his own behaviour too through self-contemplation i.e. they may not choose to
follow another individual each time (as was recommended by the roulette wheel selection operator). This
behaviour is simulated through the self-contemplation operator in the current study where if the new solution is
worse than the one which parent already had, then the parent/ grownup tries to improve themselves through self-
introspection. The self-contemplation operator thus refers to searching in the close neighbourhood of its own
current sampling space and solution; which may result in self - improvement in its own behaviour. Self-
introspection may guide the search towards a better behaviour. In context of the algorithm, this operator
provides a provision to exploit the local neighbourhood of the current solution of an individual, which might be
much better than the solutions/ behaviour of other parents. This operator simulates the self-help and self-
improvement behaviour of an individual exhibited in real life.
Similarly, every kid 0 -> I C   
 
 C   J samples variable values from within the updated
8KL 8KL
, ,

sampling interval $3= associated with every variable  += C
  

to its local neighbourhood and
further generate the updated objective function as follows:
: :
,
;KL
,
;KL

Q5?= Q ,
;KL

Q5?= Q , ,
#3 =

O"3 = AM N
"3 = @
M N
P , where #3 = C IG#3 = GJ B !
IJ
& &

Every kid adapts certain behaviour to follow and it is likely that the behaviour it chooses is worse than its
current behaviour i.e. in the optimization terminology the newly generated solution may be worse than its
current solution. In real life to tackle such situations, parents
intervention is needed who guide the children to
correct this unacceptable behaviour (this feature is simulated by behaviour correction operator in the current
study). Peer influence on kids becomes greater as they progress through their childhood and parents attempt to




correct the deteriorating behaviour of their kids. If the behaviour chosen to be followed by a kid has resulted in
its own behaviour to worsen, then the parents intervene. They supervise the adapted behaviour of their kid and if
this new behaviour is inferior then parents help their children make better behaviour choices. This process,
which is performed by the Behaviour Correction operator in SELO algorithm, is described as:
4
'6 7
The kid may sample values from any of the three sampling intervals:
/ which was formed using the kids
' 52 ' 53
own current behaviour,
/ 6 formed using current behaviour of parent
, )2
of the kid or
/ 6 formed using
current behaviour of parent , )3 of the kid. It then calculates the associated
solution or behaviour. Then one of
the better behaviours from these three will be selected based on their fitness proportions through the roulette
wheel selection. The Behaviour Correction operator as enforced by the parents will try to improve the behaviour
of the kid.
Step thus results in every family ,
< :
   = being available with, in all, < 8 =
updated behaviours (
52 53 47
parents and
 kids associated with each family B '6 C
, B '6
C, B '6 C
  :    ). The best
behaviour can be represented as
!(
(,/ :
><, =?< :    
=.

Step 5 (Exploitation)
Similar to other evolutionary algorithms, the SELO algorithm continues its iterations while exploiting better
solutions until certain stopping criterions are met. The multi-agent algorithm continues to look within the
neighbourhood of previously visited solutions till there is no significant improvement in the independent
behaviour of every family, and the difference between the individual behaviours is not very significant for
successive considerable number of iterations  i.e. if for every family
,
 < :
   =:
3 <( =+
9 <( =+&%

;


4 <( =+
9 <( =+&%

;
 
5 <( =+
9 <( =+&%

;

Values of the probabilistic parameters 0
 -
are also iteratively reduced to some minimum value (figure
). If
the families have converged in the current iteration then continue to step
, else go back to step
.

- set to a significantly small value initially 0 set to a significantly large value initially

0 0.1 1 0 0.9 1

$"
$#

F ig. 5. Representation of peer_follow_probability and family_follow_probability

Step 6
(Convergence and further search)
Once the behaviours of family members are converged (satisfying the conditions in step

), the values of
family_follow_probability 0  @  A and peer_follow_probability -  @  A are reset to their initial values so
 

 






through are then repeated again for considerable number of
learning attempts
.
The SELO algorithm is considered to be converged when the following convergence conditions are satisfied
else continue to step
a) maximum number of learning attempts .*1 is reached, 
b)  
 



 
 
   




)
Accept any of the  behaviours from current set of behaviours in the families ,
 < :    =as the final
objective function !(
(,/ value as the final solution and stop. The formulation of the SELO methodology is
explained below in detail with the algorithm flowchart in figure
.
3 Results and Discussion
This section presents the benchmark test problems used and the results and findings in order to evaluate the
performance of the proposed Social algorithm and how well it performs on finding the global optimum solution
for the unconstrained problems. The section also discusses the control parameters, precision and the stopping
criterion used for testing the optimization algorithm along with tabulated results. The performance of SELO is
compared to other widely used population based algorithms like PSO, CMAES, ABC, JDE, CLPSO, SADE,




and BSA since these algorithms are popular across scientific and real-world applications. The current algorithm
is also compared to Ideology Algorithm (IA) which is a recently proposed socio- inspired metaheuristic with
promising results. In the experiment, the computation for each of the algorithms is executed in MATLAB
R2013a on a Windows Platform with a T6400, 4 GHz Intel Core-2 Duo processor with 4 GB RAM.
3.1 Benchmar k Test Problems
To compare the relative success of the proposed algorithm with other classic population-based optimization
techniques an empirical study based on a set of  benchmark functions    (Jamil and Yang, 2013;
Yang, 2010) is conducted (refer table
). These set of test functions include a multitude of problems with
varying problem complexity levels such as unimodal, multimodal, separable and non-separable (Karaboga and
Basturk, 2008; Krink et al., 2004; Surjanovic and Bingham, 2015). Multimodal functions are used to validate
the capability of an algorithm to escape from local minimum, which tests the explorations capability of an
algorithm. Another measure of difficulty is the separability of test functions. Broadly speaking, separable
functions are easier to optimize, as compared non-separable functions, because each variable of a function is
independent of the other variables. A good blend of these test functions have been used in the current test bed
with diverse properties such as modality and separability, which proves useful in estimating and validating
various characteristics of optimization algorithms like convergence rate, precision and the general ability of
exploration and exploitation to find optimal solutions while jumping out of local minima. All benchmark test
problems are divided into four categories such as US, MS, UN, MN, and its range, formulation, characteristics
and the dimensions of these problems are listed in the table
.



 START

Initialize sampling interval reduction factor, learning attempt counter
  ,

learning iterative counter  .
Initialize parameters family_follow_probability  & peer_follow_probability 
Step 1: Initialize families
which allows the choice of whether kids follow behavior of their parents or their
peers

Initialize  families and associated individuals (parents and kids)

Each parent of every family decides to randomly follow the behaviour of a parent Step 2: Follow Behavior of Parent
from one of the other families and their associated qualities

Kids decide to follow either their parents or their siblings or kids from other families. Step 3: Follow Behavior of Kid
(Refer to Step3 of procedure)

Every individual associated with each family shrinks the sampling interval associated Step 4: Sampling-Interval Updation
with its every variable / quality

Values of parameters 
  are iteratively reduced

Step 5: Exploitation: Check for
Improvement in Behaviors across Iterations

N Family behaviours
saturated?

Y
Re-tune the parameters  and  to promote the kids to seek the help of parents
/follow the behavior of parents Step 6: After Convergence, repeat process
for more learning attempts

N
Convergence?

Y
Accept the best behavior from current set of behaviors as the final solution

STOP

F ig. 6. : SELO Algorithm Flowchart

3.2 Stopping C riterion

Each experiment has been repeated  times on each function with a random initial seed value each time.
Values below 
 are considered to be zero; considering an arithmetic precision of
 . Thus the
algorithm is assumed to have converged or terminated based on the following stopping criteria:

If the absolute value of the objective function is less than 

If the algorithm is unable to improve the quality of the found solution over a period of few preceding
iterations

Maximum number of iterations or maximum number of functions evaluations is reached

Fnc# Name Type Low Up Dim Fnc# Name Type Low Up Dim




F1 Foxholes MS -65.536 65.536 2 F26 Michalewicz2 MS 0 3.1416 2
F2 Goldstein-Price MN -2 2 2 F27 Michalewicz5 MS 0 3.1416 5
F3 Penalized MN -50 50 30 F28 Michalewicz10 MS 0 3.1416 10
F4 Penalized2 MN -50 50 30 F29 Perm MN -4 4 4
F5 Ackley MN -32 32 30 F30 Powell UN -4 5 24
F6 Beale UN -4.5 4.5 5 F31 Powersum MN 0 4 4
F7 Bohachevsky1 MS -100 100 2 F32 Quartic US -1.28 1.28 30
F8 Bohachevsky2 MN -100 100 2 F33 Rastrigin MS -5.12 5.12 30
F9 Bohachevsky3 MN -100 100 2 F34 Rosenbrock UN -30 30 30
F10 Booth MS -10 10 2 F35 Schaffer MN -100 100 2
F11 Branin MS -5 10 2 F36 Schwefel MS -500 500 30
F12 Colville UN -10 10 4 F37 Schwefel_1_2 UN -100 100 30
F13 Dixon-Price UN -10 10 30 F38 Schwefel_2_22 UN -10 10 30
F14 Easom UN -100 100 2 F39 Shekel10 MN 0 10 4
F15 Fletcher MN -3.1416 3.1416 2 F40 Shekel5 MN 0 10 4
F16 Fletcher MN -3.1416 3.1416 5 F41 Shekel7 MN 0 10 4
F17 Fletcher MN -3.1416 3.1416 10 F42 Shubert MN -10 10 2
F18 Griewank MN -600 600 30 F43 Six-hump camelback MN -5 5 2
F19 Hartman3 MN 0 1 3 F44 Sphere2 US -100 100 30
F20 Hartman6 MN 0 1 6 F45 Step2 US -100 100 30
F21 Kowalik MN -5 5 4 F46 Stepint US -5.12 5.12 5
F22 Langermann2 MN 0 10 2 F47 Sumsquares US -10 10 30
F23 Langermann5 MN 0 10 5 F48 Trid6 UN -36 36 6
F24 Langermann10 MN 0 10 10 F49 Trid10 UN -100 100 10
F25 Matyas UN -10 10 2 F50 Zakharov UN -5 10 10

Table 2 Benchmark functions used in the experiments (U-unimodal, M-multimodal, S-separable, N- non-separable, low and up
are limitations of search space)

3.3 Control Parameters

Table  summarizes the values of various control parameters used for the experimentation of proposed
algorithm SELO.

Control parameter Initial values

Maximum number of iterations (  ) 
Initial number of families created ( ) 
Number of parents in each family ( ) 
Number of children in each family ( ) 
parent_follow_probability  

follow_prob_factor_ownparent 

peer_follow_probability  

follow_prob_factor_otherkids 

sampling interval reduction factor  
 to 


Table 3 Initial value of relevant control parameters for SELO

The values of the various control parameters have been chosen based on initial trial runs carried out on the
algorithm. The parameters mentioned above have been retained as is for most of the test problems; however for
few large dimension or complex problems these parameters have been tuned to other values based on a few
initial trial runs to obtain the optimal(near optimal) solutions.

3.4 A nalysis of Results





In the study, each of the benchmark problems were solved  times and the global minimum and the running
time for each independent run or trial of the algorithm was recorded. Nature-inspired algorithms always have
some randomness owing to their stochastic nature. Thus solutions of the algorithm in the population will be
different each time the program is run; each time arriving at better or inferior solutions than they may have
arrived at during their search for newer solutions to a certain problem. To take into account this case, it is
therefore required that the overall success of a metaheuristic at solving an optimization function be considered
based on its performance across a series of trial or runs. The performance and problem solving success of
optimization algorithms are compared by using statistical measures. One algorithm may be compared to another
with the help of statistical measures to compare their correctness, algorithmic accuracy and computational
complexities. The simple statistical values like mean solution (mean), standard deviation of the mean solution
(S.D.), best solution (Best) produced and the average running time (R.T.) by the algorithms were recorded. The
detailed results are presented in table
. To validate the performance and potential of the proposed approach, the
results are then compared with some of the other popular metaheuristic algorithms like PSO, CMAES, ABC,
JDE, CLPSO, SADE BSA and IA. Table  shows the performance comparison and ranks of the algorithms in
solving the  benchmark functions. These functions provide reasonably difficult test environments for the
 .
In Table
, the first column (optimal) shows the accepted global minimum or the optimal solution to a particular
optimization function (
. The algorithms have been ranked based on their mean solutions i.e. relative
ranking based on the affectivity of the global solution attained by the algorithm for a particular  . For each
of the functions, the second row indicates the rank of each of comparator algorithms on the specified function.
The last row of Table  provides the overall ranking point (in parentheses) and consequently the final rank of
each of the algorithm. The observations from the table signify that the proposed meta-heuristic SELO could
reach the global optimum in  out of    problems during each of the runs. This record for PSO,
CMAES, ABC, JDE, CLPSO, SADE, BSA and IA is  out of  ,  out of  ,  out of

 ,  out of  ,  out of  ,  out of    ,  out of    and  out of
  respectively. It can be inferred from the current experiments and analysis that SELO and BSA are
most promising in attaining the global optimum / near optimum solutions. SELO and BSA show comparable
performances while other algorithms under comparison show relatively mediocre performances in reaching the
best minimal solutions to the test bed of simulation optimization problems. The performances of ABC, JDE,
CLPSO and IA are rather equal. It is however worth noting that all algorithms under comparison are well-
established and have reached close to the acceptable solutions for a number of test problems in the experiments;
but have not been able to rank  and reach the best solution or the global optimum in a number of the test
problems under consideration. Thus SELO and BSA prove to be the most powerful in reaching the optimum
maximum number of times, under the current study.
In assessing the overall capability of any optimization algorithm, it is important to also analyse the types of the
test problems that the algorithm solves most successfully. If the overall ranking points are to be considered
based on general performance then BSA, SADE, ABC and SELO record the most consistent and reliable
performances out of the all the algorithms, where BSA shows a superior performance over all others. CLPSO,
JDE, IA, PSO and CMAES can be ranked as average performers based on their overall ranking and problem
solving ability across all the  benchmark problems. Among all algorithms CMAES exhibits the poorest
performance recording the last place in ranking on many test functions. Taking into thought these rankings
(overall rankings as well as the places  

 

) SELO emerges as a competitive and promising
metaheuristic. Considering that SELO is a recent and a young algorithm and is still maturing, this performance
is very significant. Figure 7 illustrates the convergence plot of SELO during a run of solving the Ackley
function. It exhibits the ability of the algorithm to jump out of the local minima and reach the global minima or
the optimal solution. In the figure, we show the plot of how the behaviours () progress and steadily move
towards convergence for family number
 (comprising of total four members). The emulation and learning
behaviour of all the families is illustrated using the graphical convergence plot in figure
.
From the results of Table  it can be observed that the average performance of SELO is still acceptable
especially as compared to an advanced algorithm such as BSA (which has shown superior performance). The
average problem solving ability of SELO is better than BSA and lead on  benchmark
problems
    
 
   
   . SELO underperforms in  functions in





comparison to BSA (where it scores the highest rank of

). On functions
     
  
       SELO ranks lower than BSA.
As can be seen from the results of Table
, other than SELO none of the comparator algorithms are able to attain
the optimal global minima for Ackley and Rosenbrock. The Ackley and the Rosenbrock are non-separable
functions where interdependency is seen among the variables making them relatively difficult to solve. Both
these problems are considered as challenging for optimization problems and often used as benchmarks for
testing. Ackley becomes even more difficult since it is also multimodal. If the exploration process of an
algorithm is poorly designed, then it cannot search the function landscape effectively. The function of Ackley
poses a risk for optimization algorithms of getting stuck in one of its many local minima and the algorithm must
be able to steer the search, as far as, towards the global minima. For Rosenbrock, the global minimum lies in a
narrow valley with the shape of a parabola. However even though this valley is easy to locate; due to its non-
linearity converging to the minimum is difficult. In order to obtain consistent and good results for these
functions, the search strategy must coalesce the exploration and exploitation strategies effectively.

F ig. 7. Plot showing progress of behaviour values for Ackley function for Family-3.

F ig.8. Plot showing convergence progress pattern for all three families.





In the current study, none of the metaheuristics attain the global optimum for Schwefel, Michalewicz2 and the
noisy Quartic function; only SELO gives the closest approximations to the global solution. The Schewefel with
several local minima is particularly hard because the global minimum is at the bounds of the search space and
not in the origin as is the case with most other test problems (Molga and Smutnicki, 2005). Michalewicz
function a multimodal test function with  local minima and its complexity increases with a larger parameter
leading to a more difficult search. For the Michalewicz5 and the Michalewicz10, the proposed SELO
methodology underperforms where again all the other methodologies under comparison have been unable to
attain the expected global. The closest approximations of the global best have been recorded for BSA and ABC
with very small S.D. from the mean. The test problems for which other algorithms perform on average better
than SELO are the Penalized and Penalized2 function. SELO fairs lower in the rankings for these optimization
functions where again, most of the methodologies have been unable to attain the expected global. The Penalized
functions are challenging to solve since it composed of a combination of different sine based functions. Only
BSA and CLPSO and ABC reach the global optimum value of zero. No algorithm could meet the success
criteria in optimizing the Unimodal Dixon-Price problem. It may also be noted that SELO generally shows
comparable (or even better) success rate in solving problems with higher dimensionality. With increase in the
number of parameters or dimension, there is exponential increase in the search space. On the Powersum! "
test function, Powell! " and Langermann10! ", very few algorithms find the global optimum very
consistently, including the SELO. The flatness of the Powersum function makes it especially tricky to optimize
since the flat surface does not guide the search of the algorithm towards the optimum minima. Only SELO,
CMAES and SADE locate the global optimum (refer Table
). The non-separable function of Powell has very
diminutive minima as compared to the entire search-space making it difficult for the algorithms to locate the
global minima and SELO, CMAES and IA attain the minima on all the trials runs conducted. SELO and IA
show very consistent results on the multimodal, non-separable function Langermann ! " which is
characterized with many randomly distributed local minima, just like the Schwefel function. 
   

complexity lies in the fact that there is no implicit symmetry advantage that may simplify the optimization for
the algorithms. This indicates the strength of SELO in diversifying the search to avoid getting trapped in a local
optimum and reaching the global optimal in a speedy manner. The strong exploration capabilities of the
proposed methodology lies in the fact that even though the families in the society are all aiming for a single
goal, they are all diverse in their behaviour and attributes. A set of different families with individual members
may search the largely unknown region successively increasing the chances of arriving at the global best many
folds. This diversity in families in a social setup adds to the exploration ability of the SELO. The algorithm
spans the search space very effectively due to the diverse behaviour of different family individuals. This
exploration capability is also evident from the smaller run time of the algorithm for most of the test problems.
An analysis of the table also directs that if the local searching (exploitation) capabilities of SELO are further
tuned and enhanced, the success of the proposed methodology may be improved significantly. This is evident in
the functions where SELO ranks a second or third based on the quality of the solution and its closeness to the
global convergence value. Currently, the average performance of SELO in certain problems may be attributed to
its limitations to intensify the search in the neighbourhood of local region where the global optimum may lie.
This upcoming methodology as of now does not make use of local gradients or derivatives or the history of the
search process intensively which may give it the strength of exploiting the local information. Thus there is a
research direction that if SELO is equipped with even more local exploitation capability, the method would see
more success.
The authors also compare the mean outcomes for each of the algorithms under study by using an approximate
two-sample t test which was used for pair wise comparisons, with the statistical significance value
   . A
statistical analysis as adopted by Kashan (2014) and Zhang et al. (2008) is used to test if the proposed algorithm
performed statistically better than a comparator algorithm at solving any and every benchmark test problem
considered in the study. Table presents the results of the two-sample t test to establish the significance of
difference between the proposed SELO and the other algorithms under comparison. The following statistics are
used:




   , ! "
   
# $
#  $

 

Where 

 

are
 mean values.




are standard deviations of the results obtained by two algorithmic
approaches under comparison. The independent runs are denoted by


 . Then the degree of freedom 
is computed as under:





 
 
#  
$ ! "
   
%  
  &  %   & 
         
  
  
 

    
 
   

   

In the Table
, each of the columns (which corresponds to one comparator algorithm) an #$
sign appears in a
certain cell. This #$ sign indicates that while comparing the performance of SELO versus the corresponding
algorithm, the value of   with the corresponding  is significant at %   by our approximate two-sample t"
test. The t-tests have not been conducted for the problems where both the algorithms have obtained the optimal
solutions across all trials of a particular function. On the cases where the t-test is conducted, the results in
Table
indicate significant differences between the sample means of SELO and other competitor algorithms.
The pair-wise difference between the results achieved by SELO versus BSA is significant on  problems. For
SELO versus SADE this record is
, for SELO versus IA it is  problems, for SELO versus PSO2011 it is
 ,
for SELO versus ABC it is
. SELO is a winner over JDE on  problems; similarly this record is  for
CLPSO and  for CMAES. These results of the two-sample t-test are in agreement to the rank wise results as
presented earlier.The rank wise system was used to represent an ordering amongst the competing algorithms in
the current study. Thus in conclusion, it may be stated that SELO algorithm is very much comparable in
performance to other optimizer algorithms examined in this study.

4 Conclusions and future direction

In the paper, a new socio- inspired methodology referred to as Socio Evolution and Learning Optimization
(SELO) is proposed which mimics the natural social tendency of humans organized as family groups. It is
motivated by the evolution of social behaviour of every individual in a family. The parents and children of a
family evolve (become better) by observing and learning from one another as well as from other families. A
group of families co-existing together may be called as a society. In this societal setup, the individuals learn and
adapt to the behaviours of other individuals in the same as well as the other family individuals. Such socio-
behavioural models inspired from social tendencies in humans and their social interactions are fairly recent
developments. SELO captures the essence that that social interactions enable individuals to adapt and evolve
faster through social evolution than biological evolution based on genetic inheritance alone. Two-way evolution
is a key feature of SELO, as it allows for a two way system of learning and adaptation to take place. Evolution
and learning takes place at both: social level and population level (for every individual). When a kid is created it
is genetically similar to its parent and exhibits similar fitness; however at behaviour level may later evolve
!

!

$
  
 

  
 




 



 
$




 ! and the basic version of SELO is tested and compared with few
other accepted optimization algorithms on several benchmark problems. The results ascertain the potential and
logical correctness of the proposed work. The empirical study shows that the novel SELO algorithm shows
promising performance as measured against the other comparator algorithms. The convergence rate is
acceptable and the algorithm guides the search towards more promising areas of the solution space and the
optimal minimum in a reasonable amount of time. Since the population consists of diverse families, this
diversity lends a faster and rapid exploration capability to SELO. This can be of advantage to large scale
problems with larger sample spaces. Hence the authors anticipate that large scale problems can benefit from
SELO where the solving time can be considered to be a significant factor. Also, the intrinsic parameters to be
adjusted for SELO are comparatively less and thus SELO can be tailored to solve diverse optimization problems
with significantly less modifications and parameter tuning.
However SELO is still in its very initial stages and further efforts are needed to exploit the full effectiveness of
the methodology to attain even better problem-solving success. The results section discusses that if the local
search capability is intensified; this novel algorithm will achieve more success and provides a clear direction for
further research. The authors need to further establish the effectiveness of this metaheuristic by solving
purposeful and real world problems using this. A key task would be to identify real world problems where
advantage of adopting SELO is obvious, considering its strengths and limitations. This could also prove to be a
major direction for future research. The current version of SELO uses parameters such as
parent_follow_probability  , peer_follow_probability  , follow_prob_factor_ownparent,
follow_prob_factor_otherkids and sampling interval reduction factor  (refer table 3) which were fine tuned to


attain desired performance of the algorithm. Several preliminary trials were carried out to choose an acceptable
combination of these parameters and a strong supporting algorithm / mechanism is needed to optimize these
parameters. SELO could be designed as a self-tuning algorithm; with an adaptive feature which enables the
search algorithm to optimize and self tune the parameters at run time. Thus as the algorithm approaches
convergence, the control parameters would be iteratively reduced. This may help the algorithm exhibit expanded
local search i.e. improved exploitation. Another research direction can be to modify SELO to integrate it with
the generalized constrained handling techniques (Deshpande et al., 2013) so that it can be used to solve
constrained problems too. Popular constraint-handling techniques like penalty function and multi-objective
approaches may be considered. Most nature inspired evolutionary algorithms are adept at solving unconstrained
problems; but their performance may be impacted when attempting constrained optimization problems. Thus
designing and adaptation of evolutionary algorithms to constrained optimization is a very clear research
direction. The authors anticipate that the paper will also motivate other researchers to mature the further theory
and applications of SELO.




Table 4 Statistical solutions obtained by PSO, CMAES, ABC, JDE, CLPSO, SADE, BSA, IA and proposed SELO in the test (Mean = Mean solution; S.D. = Standard-deviation of mean solution; Best = Best
solution; R.T. = Mean Run Time in seconds)
f(x) Statistic PSO2011 CMAES ABC JDE CLPSO SADE BSA IA SELO
F1 Mean 1.3316029264876300 10.0748846367972000 0.9980038377944500 1.0641405484285200 1.8209961275956800 0.9980038377944500 0.9980038377944500 0.9980038690000000 0.9980038538690870
S.D. 0.9455237994690700 8.0277365400340800 0.0000000000000001 0.3622456829347420 1.6979175079427900 0.0000000000000000 0.0000000000000000 0.0000000000000035 0.0000013769725269
Best 0.9980038377944500 0.9980038377944500 0.9980038377944500 0.9980038377944500 0.9980038377944500 0.9980038377944500 0.9980038377944500 0.9980038685998520 0.9980038383022720
R.T. 72.527 44.788 64.976 51.101 61.650 66.633 38.125 43.535 1.750

F2 Mean 2.9999999999999200 21.8999999999995000 3.0000000465423000 2.9999999999999200 3.0000000000000700 2.9999999999999200 2.9999999999999200 3.0240147900000000 3.0013971187248700
S.D. 0.0000000000000013 32.6088098948516000 0.0000002350442161 0.0000000000000013 0.0000000000007941 0.0000000000000020 0.0000000000000011 0.0787814840000000 0.0018936009191261
Best 2.9999999999999200 2.9999999999999200 2.9999999999999200 2.9999999999999200 2.9999999999999200 2.9999999999999200 2.9999999999999200 3.0029461118668700 3.0000021202023800
R.T. 17.892 24.361 16.624 7.224 24.784 28.699 7.692 41.343 28.909
F3 Mean 0.1278728062391630 0.0241892995662904 0.0000000000000004 0.0034556340083499 0.0000000000000000 0.0034556340083499 0.0000000000000000 0.3536752140000000 0.2899597890213580
S.D. 0.2772792346028400 0.0802240262581864 0.0000000000000001 0.0189272869685522 0.0000000000000000 0.0189272869685522 0.0000000000000000 1.4205454130000000 0.0159272187796787
Best 0.0000000000000000 0.0000000000000000 0.0000000000000003 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0014898619035614 0.2497905224307240
R.T. 139.555 5.851 84.416 9.492 38.484 15.992 18.922 34.494 59.260

F4 Mean 0.0043949463343535 0.0003662455278628 0.0000000000000004 0.0007324910557256 0.0000000000000000 0.0440448539086004 0.0000000000000000 0.0179485820000000 2.3720510573781100
S.D. 0.0054747064090174 0.0020060093719584 0.0000000000000001 0.0027875840585535 0.0000000000000000 0.2227372747439610 0.0000000000000000 0.0526650620000000 0.1531241868389090
Best 0.0000000000000000 0.0000000000000000 0.0000000000000003 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000165491 2.0664368584658500
R.T. 126.507 6.158 113.937 14.367 48.667 33.019 24.309 322.808 37.105

F5 Mean 1.5214322973725000 11.7040011684582000 0.0000000000000340 0.0811017056422860 0.1863456353861950 0.7915368220335460 0.0000000000000105 0.0000000000000009 0.0000000000000000
S.D. 0.6617570384662600 9.7201961540865200 0.0000000000000035 0.3176012689149320 0.4389839299322230 0.7561593402959740 0.0000000000000034 0.0000000000000000 0.0000000000000002
Best 0.0000000000000080 0.0000000000000080 0.0000000000000293 0.0000000000000044 0.0000000000000080 0.0000000000000044 0.0000000000000080 0.0000000000000009 0.0000000000000000
R.T. 63.039 3.144 23.293 11.016 45.734 40.914 14.396 49.458 1.120

F6 Mean 0.0000000041922968 0.2540232169641050 0.0000000000000028 0.0000000000000000 0.0000444354499943 0.0000000000000000 0.0000000000000000 0.0082236060000000 0.0000997928359263
S.D. 0.0000000139615552 0.3653844307786430 0.0000000000000030 0.0000000000000000 0.0001015919507724 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0001311815541321
Best 0.0000000000000000 0.0000000000000000 0.0000000000000005 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0082236059357692 0.0000007530509495
R.T. 32.409 4.455 22.367 1.279 125.839 4.544 0.962 50.246 23.876

F7 Mean 0.0000000000000000 0.0622354533647150 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
S.D. 0.0000000000000000 0.1345061339146580 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
Best 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
R.T. 16.956 6.845 1.832 1.141 2.926 4.409 0.825 38.506 0.723

F8 Mean 0.0000000000000000 0.0072771062590204 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
S.D. 0.0000000000000000 0.0398583525142753 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
Best 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
R.T. 17.039 2.174 1.804 1.139 2.891 4.417 0.824 39.023 1.442

F9 Mean 0.0000000000000000 0.0001048363065820 0.0000000000000006 0.0000000000000000 0.0000193464326398 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
S.D. 0.0000000000000000 0.0005742120996051 0.0000000000000003 0.0000000000000000 0.0000846531630676 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000001
Best 0.0000000000000000 0.0000000000000000 0.0000000000000001 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
R.T. 17.136 2.127 21.713 1.129 33.307 4.303 0.829 40.896 3.028

F10 Mean 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0006005122443674 0.0000000000000000 0.0000000000000000 0.8346587090000000 0.0000000000000000
S.D. 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0029861918862801 0.0000000000000000 0.0000000000000000 0.0000000000000005 0.0000000000000000
Best 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.8346587086917530 0.0000000000000000
R.T. 17.072 1.375 22.395 1.099 28.508 4.371 0.790 39.978 7.373





f(x) Statistic PSO2011 CMAES ABC JDE CLPSO SADE BSA IA SELO
F11 Mean 0.3978873577297380 0.6372170283279430 0.3978873577297380 0.3978873577297380 0.3978873577297390 0.3978873577297380 0.3978873577297380 0.4156431270000000 0.3978943993817670
S.D. 0.0000000000000000 0.7302632173480510 0.0000000000000000 0.0000000000000000 0.0000000000000049 0.0000000000000000 0.0000000000000000 0.0406451050000000 0.0003536060523484
Best 0.3978873577297380 0.3978873577297380 0.3978873577297380 0.3978873577297380 0.3978873577297380 0.3978873577297380 0.3978873577297380 0.4012748152492080 0.3978822494361650
R.T. 17.049 24.643 10.941 6.814 17.283 27.981 5.450 40.099 22.72

F12 Mean 0.0000000000000000 0.0000000000000000 0.0715675060725970 0.0000000000000000 0.1593872502094070 0.0000000000000000 0.0000000000000000 0.0014898620000000 3.6688019971758100
S.D. 0.0000000000000000 0.0000000000000000 0.0579425013417103 0.0000000000000000 0.6678482786713720 0.0000000000000000 0.0000000000000000 0.0000000000000000 1.7577708967227600
Best 0.0000000000000000 0.0000000000000000 0.0013425253994745 0.0000000000000000 0.0000094069599934 0.0000000000000000 0.0000000000000000 0.0082029783984983 0.8388908577815620
R.T. 44.065 1.548 21.487 1.251 166.965 4.405 2.460 48.067 47.028

F13 Mean 0.6666666666666750 0.6666666666666670 0.0000000000000038 0.6666666666666670 0.0023282133668190 0.6666666666666670 0.6444444444444440 0.2528116640000000 0.9737369841168760
S.D. 0.0000000000000022 0.0000000000000000 0.0000000000000012 0.0000000000000002 0.0051792840882291 0.0000000000000000 0.1217161238900370 0.0000000006509080 0.0054869670667257
Best 0.6666666666666720 0.6666666666666670 0.0000000000000021 0.6666666666666670 0.0000120708732167 0.6666666666666670 0.0000000000000000 0.2528116633611470 0.9541730938494050
R.T. 167.094 3.719 37.604 18.689 216.261 47.833 21.192 67.463 38.897

F14 Mean -1.0000000000000000 -0.1000000000000000 -1.0000000000000000 -1.0000000000000000 -1.0000000000000000 -1.0000000000000000 -1.0000000000000000 -0.9997989620000000 0.0000000000000000
S.D. 0.0000000000000000 0.3051285766293650 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000167151 0.1083854312160620
Best -1.0000000000000000 -1.0000000000000000 -1.0000000000000000 -1.0000000000000000 -1.0000000000000000 -1.0000000000000000 -1.0000000000000000 -0.9997989624626810 -0.5936514558196160
R.T. 16.633 3.606 13.629 6.918 16.910 28.739 5.451 39.685 0.0129476459522760

F15 Mean 0.0000000000000000 1028.3930784026900000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
S.D. 0.0000000000000000 1298.1521820113500000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
Best 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
R.T. 27.859 15.541 40.030 2.852 4.030 6.020 2.067 38.867 0.004

F16 Mean 48.7465164446927000 1680.3460230073400000 0.0218688498331872 0.9443728655432830 81.7751618148164000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
S.D. 88.8658510972991000 2447.7484859066000000 0.0418409568792831 2.8815514827061600 379.9241117377270000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
Best 0.0000000000000000 0.0000000000000000 0.0000000000000016 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
R.T. 95.352 11.947 44.572 4.719 162.941 5.763 7.781 48.262 1.350

F17 Mean 918.9518492782850000 12340.2283326398000000 11.0681496253548000 713.7226974626920000 0.8530843976878610 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
S.D. 1652.4810858411400000 22367.1698875802000000 9.8810950146557100 1710.071307430120000 2.9208253191698800 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
Best 0.0000000000000000 0.0000000000000000 0.3274654777056860 0.0000000000000000 0.0016957837829822 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
R.T. 271.222 7.631 43.329 16.105 268.894 168.310 33.044 69.060 0.808

F18 Mean 0.0068943694819713 0.0011498935321349 0.0000000000000000 0.0048193578543185 0.0000000000000000 0.0226359326967139 0.0004930693556077 0.0000000000000000 0.0000000000000000
S.D. 0.0080565201649587 0.0036449413521107 0.0000000000000001 0.0133238235582874 0.0000000000000000 0.0283874287215679 0.0018764355751644 0.0000000000000000 0.0000000000000000
Best 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
R.T. 73.895 2.647 19.073 6.914 14.864 25.858 5.753 2.717 0.792

F19 Mean -3.8627821478207500 -3.7243887744664700 -3.8627821478207500 -3.8627821478207500 -3.8627821478207500 -3.8627821478207500 -3.8627821478207500 -3.8596352620000000 -2.2922815000937700
S.D. 0.0000000000000027 0.5407823545193820 0.0000000000000024 0.0000000000000027 0.0000000000000027 0.0000000000000027 0.0000000000000027 0.0033967610000000 0.5795350381767260
Best -3.8627821478207600 -3.8627821478207600 -3.8627821478207600 -3.8627821478207600 -3.8627821478207600 -3.8627821478207600 -3.8627821478207600 -3.8613076574052300 -3.5841184056629400
R.T. 19.280 21.881 12.613 7.509 17.504 24.804 6.009 46.167 32.987

F20 Mean -3.3180320675402500 -3.2942534432762600 -3.3219951715842400 -3.2982165473202600 -3.3219951715842400 -3.3140689634962500 -3.3219951715842400 -2.5710247593206100 -1.1719158908829300
S.D. 0.0217068148263721 0.0511458075926848 0.0000000000000014 0.0483702518391572 0.0000000000000013 0.0301641516823498 0.0000000000000013 0.0000000000000009 0.0003690446342091
Best -3.3219951715842400 -3.3219951715842400 -3.3219951715842400 -3.3219951715842400 -3.3219951715842400 -3.3219951715842400 -3.3219951715842400 -2.5710247593206100 -1.1727699585993300
R.T. 26.209 7.333 13.562 8.008 20.099 33.719 6.822 59.083 14.864

F21 Mean 0.0003074859878056 0.0064830287538208 0.0004414866359626 0.0003685318137604 0.0003100479704151 0.0003074859878056 0.0003074859878056 0.0016993410000000 0.0003493601571991
S.D. 0.0000000000000000 0.0148565973286009 0.0000568392289725 0.0002323173367683 0.0000059843325073 0.0000000000000000 0.0000000000000000 0.0000013058400000 0.0000226057336871
Best 0.0003074859878056 0.0003074859878056 0.0003230956007045 0.0003074859878056 0.0003074859941292 0.0003074859878056 0.0003074859878056 0.0016989914552560 0.0003226283751593
R.T. 84.471 13.864 20.255 7.806 156.095 45.443 11.722 48.920 15.969

F22 Mean -1.0809384421344400 -0.7323679641701760 -1.0809384421344400 -1.0764280762657400 -1.0202940450426400 -1.0809384421344400 -1.0809384421344400 -1.4315374190000000 -1.0835400071766800
S.D. 0.0000000000000006 0.4136688304155380 0.0000000000000008 0.0247042912888477 0.1190811583120530 0.0000000000000005 0.0000000000000005 0.0000000000000009 0.5277882902242550
Best -1.0809384421344400 -1.0809384421344400 -1.0809384421344400 -1.0809384421344400 -1.0809384421344400 -1.0809384421344400 -1.0809384421344400 -1.4315374193830000 -2.1933014645645000
R.T. 27.372 32.311 27.546 19.673 52.853 36.659 21.421 34.714 1.757

F23 Mean -1.3891992200744600 -0.5235864386288060 -1.4999990070800800 -1.3431399432579700 -1.4765972735526500 -1.4999992233525000 -1.4821658762555300 -1.5000000000000000 -1.4999998390866700
S.D. 0.2257194403158630 0.2585330714077300 0.0000008440502079 0.2680292304904580 0.1281777579497830 0.0000000000000009 0.0976772648082733 0.0000000000000000 0.0000000818646069
Best -1.4999992233524900 -0.7977041047646610 -1.4999992233524900 -1.4999992233524900 -1.4999992233524900 -1.4999992233524900 -1.4999992233524900 -1.5000000000000000 -1.4999999590992100
R.T. 33.809 17.940 37.986 20.333 42.488 36.037 18.930 41.848 4.708





f(x) Statistic PSO2011 CMAES ABC JDE CLPSO SADE BSA IA SELO
F24 Mean -0.9166206788680230 -0.3105071678265780 -0.8406348096500680 -0.882715279883576 -0.9431432797743700 -1.2765515661973800 -1.3127183561646500 -1.5000000000000000 -1.4999991427332700
S.D. 0.3917752367440500 0.2080317241440800 0.2000966365984320 0.3882445165494030 0.3184175870987750 0.3599594108130040 0.3158807699946290 0.0000000000000000 0.0000003717669841
Best -1.5000000000003800 -0.7976938356122860 -1.4999926800631400 -1.5000000000003800 -1.5000000000003800 -1.5000000000003800 -1.5000000000003800 -1.5000000000000000 -1.4999999303979900
R.T. 110.798 8.835 38.470 21.599 124.609 47.171 35.358 54.651 17.794

F25 Mean 0.0000000000000000 0.0000000000000000 0.0000000000000004 0.0000000000000000 0.0000041787372626 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
S.D. 0.0000000000000000 0.0000000000000000 0.0000000000000003 0.0000000000000000 0.0000161643637543 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
Best 0.0000000000000000 0.0000000000000000 0.0000000000000001 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
R.T. 25.358 1.340 19.689 1.142 31.632 4.090 0.813 35.662 1.8469296606031600

F26 Mean -1.8210436836776800 -1.7829268228561700 -1.8210436836776800 -1.8210436836776800 -1.8210436836776800 -1.8210436836776800 -1.8210436836776800 -1.8203821100000000 -1.8166465888521900
S.D. 0.0000000000000009 0.1450583631808370 0.0000000000000009 0.0000000000000009 0.0000000000000009 0.0000000000000009 0.0000000000000009 0.0000000000000014 0.0072804985619476
Best -1.8210436836776800 -1.8210436836776800 -1.8210436836776800 -1.8210436836776800 -1.8210436836776800 -1.8210436836776800 -1.8210436836776800 -1.8203821095139300 -1.8106292157333700
R.T. 19.154 26.249 17.228 9.663 18.091 28.453 7.472 34.891 1.199

F27 Mean -4.6565646397053900 -4.1008953007033700 -4.6934684519571100 -4.6893456932617100 -4.6920941990586400 -4.6884965299983800 -4.6934684519571100 -3.2820108350000000 -3.3591408962129900
S.D. 0.0557021530063238 0.4951250481844850 0.0000000000000009 0.0125797149251589 0.0075270931220834 0.0272323381095561 0.0000000000000008 0.0000000000000023 0.2009584117455920
Best -4.6934684519571100 -4.6934684519571100 -4.6934684519571100 -4.6934684519571100 -4.6934684519571100 -4.6934684519571100 -4.6934684519571100 -3.2820108345268900 -3.9631157953194900
R.T. 38.651 10.956 17.663 14.915 25.843 38.446 11.971 45.085 23.801

F28 Mean -8.9717330307549300 -7.6193507368464700 -9.6601517156413500 -9.6397230986132500 -9.6400278592589600 -9.6572038232921700 -9.6601517156413500 -6.2086254390000000 -3.9793838974626000
S.D. 0.4927013165009220 0.7904830398850970 0.0000000000000008 0.0393668145094111 0.0437935551332868 0.0105890022905617 0.0000000000000007 0.0000000000000027 0.0005104314209355
Best -9.5777818097208200 -9.1383975057875100 -9.6601517156413500 -9.6601517156413500 -9.6601517156413500 -9.6601517156413500 -9.6601517156413500 -6.2086254392105500 -3.9806353395021300
R.T. 144.093 6.959 27.051 20.803 32.801 46.395 22.250 71.652 26.425

F29 Mean 0.0119687224560441 0.0788734736114700 0.0838440014038032 0.0154105130055856 0.0198686590210374 0.0140272066690658 0.0007283694780796 1.3116221610000000 2.0169277899221400
S.D. 0.0385628598040034 0.1426911799629180 0.0778327303965192 0.0308963906374663 0.0613698943155661 0.0328868042987376 0.0014793717464195 0.5590904820000000 1.2374893392409200
Best 0.0000044608370213 0.0000000000000000 0.0129834451730589 0.0000000000000000 0.0000175219764526 0.0000000000000000 0.0000000000000000 1.0960146962658900 0.3208703882956150
R.T. 359.039 17.056 60.216 35.044 316.817 92.412 191.881 34.697 14.519

F30 Mean 0.0000130718912008 0.0000000000000000 0.0002604330013462 0.0000000000000001 0.0458769685199585 0.0000002733806735 0.0000000028443186 0.0000000000000000 0.0000000000000000
S.D. 0.0000014288348929 0.0000000000000000 0.0000394921919294 0.0000000000000002 0.0620254411839524 0.0000001788830279 0.0000000033308990 0.0000000000000000 0.0000000000000000
Best 0.0000095067504097 0.0000000000000000 0.0001682411286088 0.0000000000000000 0.0005277712020642 0.0000000944121661 0.0000000004769768 0.0000000000000000 0.0000000000000000
R.T. 567.704 14.535 215.722 194.117 252.779 360.380 144.784 153.221 1.368

F31 Mean 0.0001254882834238 0.0000000000000000 0.0077905311094958 0.0020185116261490 0.0002674563703837 0.0000000000000000 0.0000000111676630 0.0071082040000000 0.0000000000000000
S.D. 0.0001503556280087 0.0000000000000000 0.0062425841086448 0.0077448684015362 0.0003044909265796 0.0000000000000000 0.0000000184322163 0.0000000000000000 0.0000000000000000
Best 0.0000000156460198 0.0000000000000000 0.0003958766023752 0.0000000000000000 0.0000023064754605 0.0000000000000000 0.0000000000000000 0.0071082039505830 0.0000000000000000
R.T. 250.248 12.062 34.665 48.692 227.817 220.886 149.882 43.098 22.492

F32 Mean 0.0003548345513179 0.0701619169853449 0.0250163252527030 0.0013010316180679 0.0019635752485802 0.0016730768406953 0.0019955316015528 0.0002254250000000 0.0000989055208389
S.D. 0.0001410817500914 0.0288760292572957 0.0077209314806873 0.0009952078711752 0.0043423828633839 0.0007330246909835 0.0009698942217908 0.0005270410000000 0.0000521772789680
Best 0.0001014332605364 0.0299180701536354 0.0094647580732654 0.0001787238105452 0.0004206447422138 0.0005630852254632 0.0006084880639553 0.0000023800831017 0.0000104209894311
R.T. 290.669 2.154 34.982 82.124 103.283 171.637 48.237 218.722 50.737

F33 Mean 25.6367602258676000 95.9799861204982000 0.0000000000000000 1.1276202647057400 0.6301407361590880 0.8622978494808570 0.0000000000000000 0.0000000000000000 0.0000000000000000
S.D. 8.2943512684216700 56.6919245985100000 0.0000000000000000 1.0688393637536800 0.8046401822326410 0.9323785263847000 0.0000000000000000 0.0000000000000000 0.0000000000000000
Best 12.9344677422129000 29.8487565993415000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
R.T. 76.083 2.740 4.090 7.635 18.429 23.594 5.401 2.266 5.941

F34 Mean 2.6757043114269700 0.3986623855035210 0.2856833465904130 1.0630996944802500 5.7631786582751800 1.2137377447007000 0.3986623854300930 0.0000154715000000 0.0000000000000000
S.D. 12.3490058210004000 1.2164328621946200 0.6247370987465170 1.7930895051734300 13.9484817304201000 1.8518519388285700 1.2164328622195200 0.0000022373400000 0.0000000000000000
Best 0.0042535368984501 0.0000000000000000 0.0004266049929880 0.0000000000000000 0.0268003205820685 0.0001448955835246 0.0000000000000000 0.0000118803557196 0.0000000000000000
R.T. 559.966 9.462 35.865 23.278 187.894 268.449 34.681 7.250 5.855

F35 Mean 0.0000000000000000 0.4651202457398910 0.0000000000000000 0.0038863639514140 0.0019431819755029 0.0006477273251676 0.0000000000000000 0.0000000000000000 0.0000000000000000
S.D. 0.0000000000000000 0.0933685176073728 0.0000000000000000 0.0048411743884718 0.0039528023354469 0.0024650053428137 0.0000000000000000 0.0000000000000000 0.0000000000000000
Best 0.0000000000000000 0.0097159098775144 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
R.T. 18.163 24.021 7.861 4.216 8.304 5.902 1.779 33.155 1.835

F36 Mean -7684.6104757783800000 -6835.1836730901400000 -12569.4866181730000 -12304.9743375341000 -12210.8815698372000 -12549.746895737300000 -12569.486618173000000 -12569.3622100000000000 -0.3402784042291390
S.D. 745.3954005014180000 750.7338055436110000 0.0000000000022659 221.4322514436480000 205.9313376284770000 44.8939348779747000 0.0000000000024122 0.0000000273871000 3.2212919091274600
Best -8912.8855854978200000 -8340.0386911070600000 -12569.4866181730000 -12569.4866181730000 -12569.4866181730000 -12569.486618173000000 -12569.486618173000000 -12569.3622054081000000 -0.0325083488969540
R.T. 307.427 3.174 19.225 10.315 31.499 34.383 11.069 2.306 15.084





f(x) Statistic PSO2011 CMAES ABC JDE CLPSO SADE BSA IA SELO
F37 Mean 0.0000000000000000 0.0000000000000000 14.5668734126948000 0.0000000000000000 6.4655746330439100 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000009
S.D. 0.0000000000000000 0.0000000000000000 8.7128443012950300 0.0000000000000000 8.2188901353055800 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000001
Best 0.0000000000000000 0.0000000000000000 4.0427699323673400 0.0000000000000000 0.1816624029553790 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000007
R.T. 543.180 3.370 111.841 19.307 179.083 109.551 57.294 100.947 9.060

F38 Mean 0.0000000000000000 0.0000000000000000 0.0000000000000005 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.000000000000000
S.D. 0.0000000000000000 0.0000000000000000 0.0000000000000001 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.000000000000000
Best 0.0000000000000000 0.0000000000000000 0.0000000000000003 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.000000000000000
R.T. 163.188 2.558 20.588 1.494 12.563 5.627 3.208 47.009 3.520

F39 Mean -10.1061873621653000 -5.2607563471326400 -10.5364098166920000 -10.3130437162426000 -10.3130437162026000 -10.5364098166921000 -10.5364098166921000 -10.5063235800000000 -10.536281667618100
S.D. 1.6679113661236400 3.6145751818694000 0.0000000000000023 1.2234265179812200 1.2234265179736500 0.0000000000000016 0.0000000000000018 0.0000000025211900 0.0000481237097736
Best -10.5364098166921000 -10.5364098166921000 -10.5364098166920000 -10.5364098166921000 -10.5364098166920000 -10.5364098166921000 -10.5364098166920000 -10.5063235792920000 -10.5363928369535000
R.T. 31.018 11.024 16.015 8.345 37.275 28.031 7.045 55.666 12.547

F40 Mean -9.5373938082045500 -5.7308569926624600 -10.1531996790582000 -9.5656135761215700 -10.1531996790582000 -9.9847854277673500 -10.1531996790582000 -10.1529842600000000 -10.1531669871808000
S.D. 1.9062127067994200 3.5141202468383400 0.0000000000000055 1.8315977756329900 0.0000000000000076 0.9224428443735560 0.0000000000000072 0.0000000000542921 0.0000172333322304
Best -10.1531996790582000 -10.1531996790582000 -10.1531996790582000 -10.153199679058200 -10.1531996790582000 -10.1531996790582000 -10.1531996790582000 -10.1529842649756000 -10.1531973132210000
R.T. 25.237 11.177 11.958 7.947 30.885 25.569 6.864 51.507 10.116

F41 Mean -10.4029405668187000 -6.8674070870953700 -10.4029405668187000 -9.1615813354737300 -10.4029405668187000 -10.4029405668187000 -10.4029405668187000 -10.3988303400000000 -10.4028748144797000
S.D. 0.0000000000000018 3.6437803702691000 0.0000000000000006 2.8277336448396200 0.0000000000000010 0.0000000000000018 0.0000000000000017 0.0000000001978980 0.0000478046191696
Best -10.4029405668187000 -10.4029405668187000 -10.4029405668187000 -10.4029405668187000 -10.4029405668187000 -10.4029405668187000 -10.4029405668187000 -10.3988303385534000 -10.4029869270437000
R.T. 21.237 11.482 14.911 8.547 31.207 27.064 8.208 53.190 12.219

F42 Mean -186.7309073569880000 -81.5609772893002000 -186.730908831024000 -186.730908831024000 -186.730908831024000 -186.7309088310240000 -186.7309088310240000 -186.2926481000000000 -186.7153981691330000
S.D. 0.0000046401472660 66.4508342743478000 0.0000000000000236 0.0000000000000388 0.0000000000000279 0.0000000000000377 0.0000000000000224 0.0000000000000578 0.0190762312882078
Best -186.7309088310240000 -186.7309088310240000 -186.730908831024000 -186.730908831024000 -186.730908831024000 -186.7309088310240000 -186.7309088310240000 -186.2926480689880000 -186.7363874875390000
R.T. 19.770 25.225 13.342 8.213 20.344 27.109 9.002 31.766 23.870

F43 Mean -1.0316284534898800 -1.0044229658530100 -1.0316284534898800 -1.0316284534898800 -1.0316284534898800 -1.0316284534898800 -1.0316284534898800 -1.0304357800000000 -1.0303924506027700
S.D. 0.0000000000000005 0.1490105926664260 0.0000000000000005 0.0000000000000005 0.0000000000000005 0.0000000000000005 0.0000000000000005 0.0014911900000000 0.0025133845110030
Best -1.0316284534898800 -1.0316284534898800 -1.0316284534898800 -1.0316284534898800 -1.0316284534898800 -1.0316284534898800 -1.0316284534898800 -1.0314500753985900 -1.0314918740874000
R.T. 16.754 24.798 11.309 7.147 18.564 27.650 5.691 39.897 2.245

F44 Mean 0.0000000000000000 0.0000000000000000 0.0000000000000004 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.000000000000000
S.D. 0.0000000000000000 0.0000000000000000 0.0000000000000001 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.000000000000000
Best 0.0000000000000000 0.0000000000000000 0.0000000000000003 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.000000000000000
R.T. 159.904 2.321 21.924 1.424 14.389 5.920 3.302 174.577 1.200

F45 Mean 2.3000000000000000 0.0666666666666667 0.0000000000000000 0.9000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000538870000000 0.0000000000000010
S.D. 1.8597367258983700 0.2537081317024630 0.0000000000000000 3.0211895350832500 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000005399890 0.0000000000000001
Best 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000538860819891 0.0000000000000006
R.T. 57.276 1.477 1.782 2.919 3.042 4.307 0.883 2.215 5.274

F46 Mean 0.1333333333333330 0.2666666666666670 0.0000000000000000 0.0000000000000000 0.2000000000000000 0.0000000000000000 0.0000000000000000 -0.0153463301609662 0.0000000000000000
S.D. 0.3457459036417600 0.9444331755018490 0.0000000000000000 0.0000000000000000 0.4068381021724860 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
Best 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 -0.0153463301609662 0.0000000000000000
R.T. 20.381 2.442 1.700 1.074 6.142 4.319 0.764 31.068 0.230

F47 Mean 0.0000000000000000 0.0000000000000000 0.0000000000000005 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
S.D. 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
Best 0.0000000000000000 0.0000000000000000 0.0000000000000003 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
R.T. 564.178 2.565 24.172 1.870 15.948 6.383 4.309 31.296 1.009

F48 Mean -50.0000000000002000 -50.0000000000002000 -49.9999999999997000 -50.0000000000002000 -49.4789234062579000 -50.0000000000002000 -50.0000000000002000 -44.7416748700000000 -46.672022811734100
S.D. 0.0000000000000361 0.0000000000000268 0.0000000000001408 0.0000000000000354 1.3150773145311700 0.0000000000000268 0.0000000000000361 0.0000000000000217 1.1721159591339900
Best -50.0000000000002000 -50.0000000000002000 -50.0000000000001000 -50.0000000000002000 -49.9999994167392000 -50.0000000000002000 -50.0000000000002000 -44.7416748706606000 -48.933141750726300
R.T. 24.627 8.337 22.480 8.623 142.106 36.804 7.747 52.486 30.015

F49 Mean -210.000000000001000 -210.000000000003000 -209.999999999947000 -210.000000000003000 -199.592588547503000 -210.000000000003000 -210.000000000003000 -150.5540859185450000 -162.571266865506000
S.D. 0.0000000000009434 0.0000000000003702 0.0000000000138503 0.0000000000008251 9.6415263953591700 0.0000000000004625 0.0000000000003950 0.0000000000000000 0.2649613601835890
Best -210.000000000003000 -210.000000000003000 -209.999999999969000 -210.000000000004000 -209.985867409029000 -210.000000000004000 -210.000000000004000 -150.5540859185450000 -162.922114827822000
R.T. 48.580 5.988 36.639 11.319 187.787 54.421 11.158 70.887 39.873





f(x) Statistic PSO2011 CMAES ABC JDE CLPSO SADE BSA IA SELO
F50 Mean 0.0000000000000000 0.0000000000000000 0.0000000402380424 0.0000000000000000 0.0000000001597805 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
S.D. 0.0000000000000000 0.0000000000000000 0.0000002203520334 0.0000000000000000 0.0000000006266641 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
Best 0.0000000000000000 0.0000000000000000 0.0000000000000210 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
R.T. 86.369 1.868 86.449 1.412 157.838 4.930 5.702 33.573 3.261

Table 4 continued...Statistical solutions obtained by PSO, CMAES, ABC, JDE, CLPSO, SADE, BSA, IA and proposed SELO in the test (Mean = Mean solution; S.D. = Standard-deviation of mean solution; Best =
Best solution; R.T. = Mean Run Time in seconds)

Table 5 Performance comparison and rankings of algorithms PSO, CMAES, ABC, JDE, CLPSO, SADE, BSA, IA and proposed SELO on 50 benchmark functions
F(x) Optimal Statistics PSO2011 CMAES ABC JDE CLPSO SADE BSA IA SELO
F1 0.998003 Mean 1.3316029264876300 10.0748846367972000 0.9980038377944500 1.0641405484285200 1.8209961275956800 0.9980038377944500 0.9980038377944500 0.9980038690000000 0.9980038538690870
(+) (+) (+) (+)
Rank 3 5 1 2 4 1 1 1 1

F2 3 Mean 2.9999999999999200 21.8999999999995000 3.0000000465423000 2.9999999999999200 3.0000000000000700 2.9999999999999200 2.9999999999999200 3.0240147900000000 3.0013971187248700
(+) (+)
Rank 1 3 1 1 1 1 1 2 1

F3 0 Mean 0.1278728062391630 0.0241892995662904 0.0000000000000004 0.0034556340083499 0.0000000000000000 0.0034556340083499 0.0000000000000000 0.3536752140000000 0.2899597890213580

Rank 5 4 2 3 1 3 1 7 6

F4 0 Mean 0.0043949463343535 0.0003662455278628 0.0000000000000004 0.0007324910557256 0.0000000000000000 0.0440448539086004 0.0000000000000000 0.0179485820000000 2.3720510573781100

Rank 5 3 2 4 1 7 1 6 8

F5 0 Mean 1.5214322973725000 11.7040011684582000 0.0000000000000340 0.0811017056422860 0.1863456353861950 0.7915368220335460 0.0000000000000105 0.0000000000000009 0.0000000000000000
(+) (+) (+) (+) (+) (+) (+) (+)
Rank 8 9 4 5 6 7 3 2 1

F6 0 Mean 0.0000000041922968 0.2540232169641050 0.0000000000000028 0.0000000000000000 0.0000444354499943 0.0000000000000000 0.0000000000000000 0.0082236060000000 0.0000997928359263

Rank 3 7 2 1 4 1 1 6 5

F7 0 Mean 0.0000000000000000 0.0622354533647150 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
(+)
Rank 1 2 1 1 1 1 1 1 1

F8 0 Mean 0.0000000000000000 0.0072771062590204 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
(+)
Rank 1 2 1 1 1 1 1 1 1

F9 0 Mean 0.0000000000000000 0.0001048363065820 0.0000000000000006 0.0000000000000000 0.0000193464326398 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
(+) (+) (+)
Rank 1 4 2 1 3 1 1 1 1

F10 0 Mean 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0006005122443674 0.0000000000000000 0.0000000000000000 0.8346587090000000 0.0000000000000000
(+) (+)
Rank 1 1 1 1 2 1 1 3 1

F11 0.39788 Mean 0.3978873577297380 0.6372170283279430 0.3978873577297380 0.3978873577297380 0.3978873577297390 0.3978873577297380 0.3978873577297380 0.4156431270000000 0.3978943993817670
(+) (+)
Rank 1 3 1 1 1 1 1 2 1

F12 0 Mean 0.0000000000000000 0.0000000000000000 0.0715675060725970 0.0000000000000000 0.1593872502094070 0.0000000000000000 0.0000000000000000 0.0014898620000000 3.6688019971758100

Rank 1 1 3 1 4 1 1 2 5





F(x) Optimal Statistics PSO2011 CMAES ABC JDE CLPSO SADE BSA IA SELO

F13 0 Mean 0.6666666666666750 0.6666666666666670 0.0000000000000038 0.6666666666666670 0.0023282133668190 0.6666666666666670 0.6444444444444440 0.2528116640000000 0.9737369841168760

Rank 6 5 1 5 2 5 4 3 7
F14 -1 Mean -1.0000000000000000 -0.1000000000000000 -1.000000000000000 -1.000000000000000 -1.000000000000000 -1.000000000000000 -1.000000000000000 -0.999798962000000 0.0000000000000000

Rank 1 3 1 1 1 1 1 2 4

F15 0 Mean 0.0000000000000000 1028.3930784026900000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
(+)
Rank 1 2 1 1 1 1 1 1 1

F16 0 Mean 48.7465164446927000 1680.3460230073400000 0.0218688498331872 0.9443728655432830 81.7751618148164000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
(+) (+) (+) (+) (+)
Rank 4 6 2 3 5 1 1 1 1

F17 0 Mean 918.9518492782850000 12340.2283326398000000 11.0681496253548000 713.7226974626920000 0.8530843976878610 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
(+) (+) (+) (+) (+)
Rank 5 6 3 4 2 1 1 1 1

F18 0 Mean 0.0068943694819713 0.0011498935321349 0.0000000000000000 0.0048193578543185 0.0000000000000000 0.0226359326967139 0.0004930693556077 0.0000000000000000 0.0000000000000000
(+) (+) (+) (+) (+)
Rank 5 3 1 4 1 6 2 1 1

F19 -3.86278 Mean -3.8627821478207500 -3.7243887744664700 -3.8627821478207500 -3.8627821478207500 -3.8627821478207500 -3.8627821478207500 -3.8627821478207500 -3.8596352620000000 -2.2922815000937700

Rank 1 3 1 1 1 1 1 2 4

F20 -3.32237 Mean -3.3180320675402500 -3.2942534432762600 -3.3219951715842400 -3.2982165473202600 -3.3219951715842400 -3.3140689634962500 -3.3219951715842400 -2.5710247593206100 -1.1719158908829300

Rank 2 5 1 4 1 3 1 6 7

F21 0.0003 Mean 0.0003074859878056 0.0064830287538208 0.0004414866359626 0.0003685318137604 0.0003100479704151 0.0003074859878056 0.0003074859878056 0.0016993410000000 0.0003493601571991
(+) (+) (+) (+)
Rank 1 4 2 2 1 1 1 3 1

F22 -1.0809 Mean -1.0809384421344400 -0.7323679641701760 -1.0809384421344400 -1.0764280762657400 -1.0202940450426400 -1.0809384421344400 -1.0809384421344400 -1.4315374190000000 -1.0835400071766800
(+) (+) (+) (+)
Rank 1 5 1 3 4 1 1 6 1

F23 -1.5 Mean -1.3891992200744600 -0.5235864386288060 -1.4999990070800800 -1.3431399432579700 -1.4765972735526500 -1.4999992233525000 -1.4821658762555300 -1.5000000000000000 -1.4999998390866700
(+) (+) (+) (+) (+)
Rank 4 6 1 5 3 1 2 1 1

F24 -1.5 Mean -0.9166206788680230 -0.3105071678265780 -0.8406348096500680 -0.8827152798835760 -0.9431432797743700 -1.2765515661973800 -1.3127183561646500 -1.5000000000000000 -1.4999991427332700
(+) (+) (+) (+) (+) (+) (+)
Rank 5 8 7 6 4 3 2 1 1

F25 0 Mean 0.0000000000000000 0.0000000000000000 0.0000000000000004 0.0000000000000000 0.0000041787372626 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
(+) (+)
Rank 1 1 2 1 3 1 1 1 1

F26 -1.8013 Mean -1.8210436836776800 -1.7829268228561700 -1.8210436836776800 -1.8210436836776800 -1.8210436836776800 -1.8210436836776800 -1.8210436836776800 -1.8203821100000000 -1.8166465888521900
(+) (+) (+) (+) (+) (+) (+) (+)
Rank 4 2 4 4 4 4 4 3 1

F27 -4.6876 Mean -4.6565646397053900 -4.1008953007033700 -4.6934684519571100 -4.6893456932617100 -4.6920941990586400 -4.6884965299983800 -4.6934684519571100 -3.2820108350000000 -3.3591408962129900

Rank 5 6 4 2 3 1 4 8 7

F28 -9.66015 Mean -8.9717330307549300 -7.6193507368464700 -9.6601517156413500 -9.6397230986132500 -9.6400278592589600 -9.6572038232921700 -9.6601517156413500 -6.2086254390000000 -3.9793838974626000

Rank 5 6 1 4 3 2 1 7 8





F(x) Optimal Statistics PSO2011 CMAES ABC JDE CLPSO SADE BSA IA SELO

F29 0 Mean 0.0119687224560441 0.0788734736114700 0.0838440014038032 0.0154105130055856 0.0198686590210374 0.0140272066690658 0.0007283694780796 1.3116221610000000 2.0169277899221400

Rank 2 6 7 4 5 3 1 8 9

F30 0 Mean 0.0000130718912008 0.0000000000000000 0.0002604330013462 0.0000000000000001 0.0458769685199585 0.0000002733806735 0.0000000028443186 0.0000000000000000 0.0000000000000000
(+) (+) (+) (+) (+) (+)
Rank 5 1 6 2 7 4 3 1 1

F31 0 Mean 0.0001254882834238 0.0000000000000000 0.0077905311094958 0.0020185116261490 0.0002674563703837 0.0000000000000000 0.0000000111676630 0.0071082040000000 0.0000000000000000
(+) (+) (+) (+) (+) (+)
Rank 3 1 7 5 4 1 2 6 1

F32 0 Mean 0.0003548345513179 0.0701619169853449 0.0250163252527030 0.0013010316180679 0.0019635752485802 0.0016730768406953 0.0019955316015528 0.0002254250000000 0.0000989055208389
(+) (+) (+) (+) (+) (+) (+) (+)
Rank 3 9 8 4 6 5 7 2 1

F33 0 Mean 25.6367602258676000 95.9799861204982000 0.0000000000000000 1.1276202647057400 0.6301407361590880 0.8622978494808570 0.0000000000000000 0.0000000000000000 0.0000000000000000
(+) (+) (+) (+) (+)
Rank 5 6 1 4 2 3 1 1 1

F34 0 Mean 2.6757043114269700 0.3986623855035210 0.2856833465904130 1.0630996944802500 5.7631786582751800 1.2137377447007000 0.3986623854300930 0.0000154715000000 0.0000000000000000
(+) (+) (+) (+) (+) (+) (+) (+)
Rank 8 5 3 6 9 7 4 2 1

F35 0 Mean 0.0000000000000000 0.4651202457398910 0.0000000000000000 0.0038863639514140 0.0019431819755029 0.0006477273251676 0.0000000000000000 0.0000000000000000 0.0000000000000000
(+) (+) (+) (+)
Rank 1 5 1 4 3 2 1 1 1

F36 0 Mean -7684.6104757783800000 -6835.1836730901400000 - - - - - - -0.3402784042291390

(+) (+) 12569.4866181730000000 12304.9743375341000000 12210.8815698372000000 12549.7468957373000000 12569.4866181730000000 12569.3622100000000000
(+) (+) (+) (+) (+) (+)
Rank 3 2 8 5 4 6 8 7 1

F37 0 Mean 0.0000000000000000 0.0000000000000000 14.5668734126948000 0.0000000000000000 6.4655746330439100 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
(+) (+)
Rank 1 1 4 1 3 1 1 1 1

F38 0 Mean 0.0000000000000000 0.0000000000000000 0.0000000000000005 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
(+)
Rank 1 1 2 1 1 1 1 1 1

F39 -10.5364 Mean -10.1061873621653000 -5.2607563471326400 -10.5364098166920000 -10.3130437162426000 -10.3130437162026000 -10.5364098166921000 -10.5364098166921000 -10.5063235800000000 -10.5362816676181000
(+) (+) (+) (+) (+)
Rank 4 5 1 3 3 1 1 2 1

F40 -10.1532 Mean -9.5373938082045500 -5.7308569926624600 -10.1531996790582000 -9.5656135761215700 -10.1531996790582000 -9.9847854277673500 -10.1531996790582000 -10.1529842600000000 -10.1531669871808000
(+) (+) (+) (+)
Rank 4 5 1 3 1 2 1 1 1
F41 -10.4029 Mean -10.4029405668187000 -6.8674070870953700 -10.4029405668187000 -9.1615813354737300 -10.4029405668187000 -10.4029405668187000 -10.4029405668187000 -10.3988303400000000 -10.4028748144797000
(+) (+) (+)
Rank 1 4 1 3 1 1 1 2 1

F42 - Mean -186.7309073569880000 -81.5609772893002000 -186.7309088310240000 -186.7309088310240000 -186.7309088310240000 -186.7309088310240000 -186.7309088310240000 -186.2926481000000000 -
186.7309 186.7153981691330000
Rank 1 4 1 1 1 1 1 3 2

F43 -1.03016 Mean -1.031628453489880 -1.0044229658530100 -1.0316284534898800 -1.0316284534898800 -1.0316284534898800 -1.0316284534898800 -1.0316284534898800 -1.0304357800000000 -1.0303924506027700
(+) (+) (+) (+) (+) (+) (+)
Rank 2 3 2 2 2 2 2 1 1





F(x) Optimal Statistics PSO2011 CMAES ABC JDE CLPSO SADE BSA IA SELO

F44 0 Mean 0.0000000000000000 0.0000000000000000 0.0000000000000004 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
(+)
Rank 1 1 2 1 1 1 1 1 1

F45 0 Mean 2.3000000000000000 0.0666666666666667 0.0000000000000000 0.9000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000538870000000 0.0000000000000010

Rank 6 4 1 5 1 1 1 3 2

F46 0 Mean 0.1333333333333330 0.2666666666666670 0.0000000000000000 0.0000000000000000 0.2000000000000000 0.0000000000000000 0.0000000000000000 -0.015346330160966 0.0000000000000000
(+) (+) (+) (+)
Rank 3 5 1 1 4 1 1 2 1

F47 0 Mean 0.0000000000000000 0.0000000000000000 0.0000000000000005 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
(+)
Rank 1 1 2 1 1 1 1 1 1

F48 -50 Mean -50.0000000000002000 -50.0000000000002000 -49.9999999999997000 -50.0000000000002000 -49.4789234062579000 -50.0000000000002000 -50.0000000000002000 -44.7416748700000000 -46.6720228117341000

Rank 1 1 1 1 2 1 1 4 3

F49 -200 Mean - -210.0000000000030000 -209.9999999999470000 -210.0000000000030000 -199.5925885475030000 -210.0000000000030000 -210.0000000000030000 -150.5540859185450000 -
210.0000000000010000 162.5712668655060000
Rank 3 4 2 4 1 4 4 6 5

F50 0 Mean 0.0000000000000000 0.0000000000000000 0.0000000402380424 0.0000000000000000 0.0000000001597805 0.0000000000000000 0.0000000000000000 0.0000000000000000 0.0000000000000000
(+) (+)
Rank 1 1 3 1 2 1 1 1 1

Rank (Point) 5(22) 8(11) 4(24) 7(20) 7(20) 3(31) 1(36) 6(21) 2(35)

Table 5 Performance comparison and rankings of algorithms PSO, CMAES, ABC, JDE, CLPSO, SADE, BSA, IA and proposed SELO on 50 benchmark functions
*

 
  
 
 


  


with other algorithm, value of  with the corresponding  is significant at  
 by the two-sample ttest





References
1. Ahmadi-Javid, A. (2011) 2Anarchic society optimization: A human-inspired method3
%)
Evolutionary
Computation (CEC), 2011 IEEE Congress, IEEE, New Orleans, USA, pp.2586-2592.
2. Atashpaz-Gargari, E., & Lucas, C. (2007) 2Imperialist competitive algorithm: an algorithm for
optimization inspired by imperialistic competition3
%n Evolutionary computation (CEC), 2007 IEEE
Congress, IEEE, Singapore, pp.4661-4667.
3. Bandura, A. (1962) 2Social learning through imitation3
%n M. R. Jones (Ed.), Nebraska Symposium on
Motivation. Lincoln: University of Nebraska Press.
4. Bandura, A., & Walters, R. H. (1977) Social learning theory, General Learning Press, New York.
5. Biswas, A., Mishra, K. K., Tiwari, S., & Misra, A. K. (2013) 2Physics-inspired optimization
algorithms: a survey3
Journal of Optimization, Vol. 2013, Article ID 438152, 16 pages,
doi:10.1155/2013/438152.
6. Bonabeau, E., Dorigo, M., & Theraulaz, G. (1999) Swarm intelligence: from natural to artificial
systems, No. 1, Oxford university press Santa Fe, USA.
7. Brest, J., Greiner, S., Boskovic, B., Mernik, M., & Zumer, V. (2006), 2Self-adapting control parameters
in differential evolution: A comparative study on numerical benchmark problems3
IEEE transactions
on evolutionary computation, Vol. 10 No. 6, pp.646-657.
8. Brooks, S. P., & Morgan, B. J. (1995) 2Optimization using simulated annealing3
The Statistician, Vol.
44 No. 2, pp.241-257. doi: 10.2307/2348448
9. Brownlee, J. (2011) Clever algorithms: nature-inspired programming recipes, Jason Brownlee,
Australia.
10. Civic%*#'/


!- *&

  
2A conceptual comparison of the Cuckoo-search, particle swarm
optimization, differential evolution and artificial bee colony algorithms3
Artificial intelligence review,
Vol. 39 No. 315, pp.1-32. doi:10.1007/s10462-011-9276-0
11. Civicioglu, P. (2013) 2Backtracking search optimization algorithm for numerical optimization
problems3
Applied Mathematics and Computation, Vol. 219 No. 15, pp.8121-8144.
12. Clerc, Maurice (2012) Standard particle swarm optimization, 15 pages <hal-00764996>.
13. !-$+) !


$.))%



/'&,)%


  
2*)-.,%).
$) '%)#
%)
"%,!"'0
'#*,%.$(3

in Cybernetics (CYBCONF), 2013 IEEE International Conference, IEEE, United States, pp. 186-190.
14. *,%#*

%,..,%


./.1'!

  
2 ).
*'*)0
*+.%(%1.%*)3 IEEE computational
intelligence magazine, Vol. 1 No. 4, pp.28-39.
15. Emam%


!,&$-$)

  
2Election algorithm: a new socio-politically inspired strategy3
AI
Communications, Vol. 28 No. 3, pp.591-603.
16. Eisenberg, M. (2008) 2The Peer Assumption: A review of the nurture --/(+.%*)3
the journal of the
learning sciences, Vol.17 No. 4, pp.588-594. http://dx.doi.org/10.1080/10508400802394906.
17. Filho, C., de Lima Neto, F., Lins, A., Nascimento, A., & Lima, M. (2009) 2Fish school search3
in
Nature-inspired algorithms for optimisation, Springer Berlin Heidelberg, Vol. 193, pp. 261-277.
18. Fister Jr, I., Yang, X. S., Fister, I., Brest, J., & Fister, D. (2013) A Brief Review of Nature-Inspired
Algorithms for Optimization. arXiv preprint arXiv:1307.4186.
19. Geem, Z. W. (2010) 2State-of-the-art in the structure of harmony search algorithm3
in Recent advances
in harmony search algorithm, Studies in computational intelligence, Springer Berlin Heidelberg Vol.
270, pp.1-10.
20. Gendreau, M., & Potvin, J. Y. (2010) Handbook of metaheuristics, New York: Springer, Vol. 2.
21. Goffe, W. L., Ferrier, G. D., & Rogers, J. (1994) 2Global optimization of statistical functions with
simulated annealing3
Journal of econometrics, Vol. 60 No. 1-2, pp.65-99.
22. Goldberg, D. E., & Deb, K. (1991) 2A comparative analysis of selection schemes used in genetic
algorithms3
in Foundations of genetic algorithms, Morgan Kaufmann Publishers Inc., 1, pp.69-93.
23. Goldsmith, Edward (1978) The stable society: its structure and control: towards a social cybernetics.
Wadebridge Press, Wadebridge, Eng.
24. )-!)

  
2The CMA evolution strategy: a comparing review3
%n Towards a new evolutionary
computation, Studies in Fuzziness and Soft Computing , vol. 192, Springer, pp. 75-102
25. Hassanien, A. E., & Emary, E. (2016) Swarm intelligence: principles, advances, and applications. CRC
Press.
26. Hechter, M., & Horne, C. (2009) Theories of Social Order: A Reader, 2nd edition. Stanford University
Press, Stanford, CA.
27. Ho, 


!+0)!


  
2Simple explanation of the no-free-lunch theorem and its
implications3
Journal of optimization theory and applications, Vol. 115 No. 3, pp.549-570.
28. Hosseini, S., & Al Khaled, A. (2014) 2A survey on the imperialist competitive algorithm metaheuristic:
Implementation in engineering domain and directions for future research3
Applied Soft
Computing, Vol. 24, pp.1078-1094.



 29. Huan, T.T., Kulkarni, A.J., Kanesan J. et al. (2016) <Ideology algorithm: a socio-inspired optimization
methodology=
Neural Computing & Applications, pp.1-32. Doi: 10.1007/s00521-016-2379-4.
30. Igel, C., Hansen, N., & Roth, S. (2007) <Covariance matrix adaptation for multi-objective
optimization=
Evolutionary computation, Vol. 15 No. 1, pp.1-28.
31. Jamil, M., & Yang, X. S. (2013) <A literature survey of benchmark functions for global optimisation
problems=
International Journal of Mathematical Modelling and Numerical Optimisation, Vol. 4 No. 2,
pp.150-194.
32. $3$%1*$


$4563-

 
<A powerful and efficient algorithm for numerical function
optimization: artificial bee colony (ABC) algorithm=, J Global Optim, Vol. 39, pp.459;471.
33. Karaboga, D., & Basturk, B. (2008) <On the performance of artificial bee colony (ABC) algorithm=

Applied soft computing, Vol. 8 No. 1, pp.687-697.
34. $4+$0


 
<League championship algorithm: a new algorithm for numerical function
optimization=
,n International conference on soft computing and pattern recognition (SOCPAR'09),
IEEE, Singapore, pp.43-48.
35. Kashan, A. H. (2011) <An efficient algorithm for constrained global optimization and application to
mechanical engineering design: League championship algorithm (LCA)=
Computer-Aided
Design, Vol. 43 No. 12, pp.1769-1792.
36. Kashan, A. H. (2014) <League Championship Algorithm (LCA): An algorithm for global optimization
inspired by sport championships=
Applied Soft Computing, Vol. 16, pp.171-200.
37. (00('9


%(3+$35

 
<$35,&.(
48$3/
125,/,:$5,10=
,0
31&((',0*4
1)

05(30$5,10$.

Conference on Neural Networks, IEEE, Vol. 4, pp.1942-1948.
38. Kirkpatrick, S., Gelatt, C. D., & Vecchi, M. P. (1983) Optimization by simulated annealing. Science,
220(4598), 671-680.
39. Koppen, M., Wolpert, D. H., & Macready, W. G. (2001) <Remarks on a recent paper on the" no free
lunch" theorems=
IEEE Transactions on Evolutionary Computation, Vol. 5 No. 3, pp.295-296.
40. Kri0-
"
,.,2,&


1*(.


  
<1,49
125,/,:$5,10
231%.(/4-a particular challenge for
',))(3(05,$.
(71.65,10=
,0 Congress on Evolutionary Computation(CEC2004), IEEE , Vol. 1, pp.332-
339.
41. Ku.-$30,



"$,

 
<Probability collectives: a multi-agent approach for solving
comb,0$513,$.
125,/,:$5,10
231%.(/4=
Applied Soft Computing, Vol. 10 No. 3, pp.759-771.
42. 6.-$30,


636*-$3



6/$3

  
<1+135
05(..,*(0&(
a self supervised learning
behavior=
,0 Systems, Man, and Cybernetics (SMC), IEEE International Conference, IEEE,
Manchester, UK, pp. 1396-1400
43. Kulkarni, A. J., & Shabi3

  
<Solving 0;1 knapsack problem using cohort intelligence
algorithm=
International Journal of Machine Learning and Cybernetics, Vol. 7 No.3, pp.427-441.
44. Kulkarni, A. J., Krishnasamy, G., & Abraham, A. (2016) Cohort Intelligence: A Socio-Inspired
Optimization Method, Intelligent systems reference library, Springer, Vol. 114, pp.1-134.
45. 6/$3

  
<Blending roulette wheel selection & rank selection in genetic algorithms=

International Journal of Machine Learning and Computing, Vol. 2 No. 4, p.365.
46. Kuo, H. C., & Lin, C. H. (2013) <Cultural evolution algorithm for global optimizations and its
applications=
Journal of applied research and technology, Vol. 11 No. 4, pp.510-522.
47. Lam AYS, Li VOK (2010) <Chemical-reaction-inspired metaheuristic for optimization=

"3$04

Evol Comput, Vol. 14 No. 3, pp.381;399.
48. Liang, J. J., Qin, A. K., Suganth$0



$4-$3
!
  
<Comprehensive learning particle swarm
optimizer for global optimization of multimodal functions= IEEE transactions on evolutionary
computation, Vol.10 No. 3, pp. 281-295.
49. Liu, Z. Z., Chu, D. H., Song, C., Xue, X., & Lu, B. Y. (2016) <Social learning optimization (SLO)
algorithm paradigm and its application in QoS-aware cloud service composition=
Information
Sciences, Vol. 326, pp.315-333.
50. Luke, S. (2013) Essentials of Metaheuristics, Lulu, second edition, available for free at
http://cs.gmu.edu/~sean/book/metaheuristics/
51. Lv, W., He, C., Li, D., Cheng, S., Luo, S., & Zhang, X. (2010) <Election campaign optimization
algorithm=
,0
Procedia Computer Science, Vol. 1 No. 1, pp.1377-1386.
52. $&&1%9


 
<The role of parents in the socialization of children: An historical 17(37,(8= in
Developmental psychology, Vol. 28 No. 6, pp.1006-1017.doi: 10.1037/0012-1649.28.6.1006.
53. Ma3,0,


#$.&:$-

  
<Particle swarm optimization (PSO). A tutorial=, Chemometrics and
Intelligent Laboratory Systems, Vol.149, pp.153-165.
54. Molga, M., & Smutnicki, C. (2005) Test functions for optimization needs, Test functions for
optimization needs, pp.101.





 55. Moosavian


**"-!,&


  
4Soccer league competition algorithm: A novel meta-
heuristic algorithm for optimal design of water distribution networks5
Swarm and Evolutionary
Computation, Vol. 17, pp.14-24.
56. **-!0&!)

  
4Soccer league competition algorithm for solving knapsack problems5
Swarm
and Evolutionary Computation, Vol. 20, pp.14-22.
57. Nanda, S. J., & Panda, G. (2014) 4A survey on nature inspired metaheuristic algorithms for partitional
clustering5
Swarm and Evolutionary computation, Vol. 16, pp.1-18.
58. N#,&


&,,*)#)

 
4Recent advances in differential evolution: a survey and experimental
!)!(2-&-5
Artificial Intelligence Review, Vol. 33 No.1-2, pp.61-106.
59. Omran MGH, Clerc M (2011) https://www.particleswarm.info/Programs.html (Accessed 15 Dec 2016)
60. Pencheva, T., Atanassov, K., & Shannon, A. (2009) 4Modelling of a roulette wheel selection operator
in genetic algorithms using generalized nets5
Int. J. Bioautomation, Vol. 13 No. 4, pp.257-264.
61. Qin, A. Kai, and P*))/.%/,!&

/$!).%!)
  
4Self-adaptive differential evolution algorithm for
numerical optimization5
&) Evolutionary Computation (CEC), 2005 IEEE Congress, IEEE, Edinburgh,
UK, Vol. 2, pp.1785-1791.
62. Rao, R. V., Savsani, V


!'%!,&!


  
4Teaching3learning-based optimization: an
optimization method for continuous non-linear large scale problems5
Information Sciences, Vol. 183
No. 1, pp.1-15.
63. !2


&#1


  
4Society and civilization: An optimization algorithm based on the
simulation of social behavior5
IEEE Transactions on Evolutionary Computation, Vol. 7 No. 4, pp.386-
396.
64. Reynolds



0#,"(&'

 
4Problem solving using cultural algorithms5
&n Evolutionary
Computation, IEEE World Congress on Computational Intelligence, Proceedings of the First IEEE
Conference, Orlando, FL, USA, pp. 645-650.
65. Sa.!+!.%2


!&'

  
4Social group optimization (SGO): a new population evolutionary
optimization technique5
Complex & Intelligent Systems, Vol. 2 No. 3, pp.173-203.
66. Storn, R., & Price, K. (1997) 4Differential evolution3a simple and efficient heuristic for global
optimization over continuous spaces5
Journal of global optimization, Vol. 11 No. 4, pp.341-359.
67. Surjanovic, S., & Bingham, D.(2015).British Columbia https://www.sfu.ca/~ssurjano/optimization.html
(Accessed 15 Jan 2017)
68. Talbi, E. G. (2009) Metaheuristics: from design to implementation, Vol. 74, John Wiley & Sons.
69. Wolpert, D. H., & Macready, W. G. (1997) 4No free lunch theorems for optimization5
IEEE
transactions on evolutionary computation, Vol. 1 No. 1, pp.67-82.
70. Xie, Q., Lv, W., Liu, Z., Zhang, X., Luo, S., & Cheng, S. (2010) 4Constrained optimization with
Election campaign algorithm5
&n Industrial Mechatronics and Automation (ICIMA), 2nd International
Conference, IEEE, Wuhan, China, Vol. 1, pp. 370-373.
71. Xu, Y., Cui, Z., & Zeng, J. (2010) 4Social emotional optimization algorithm for nonlinear constrained
optimization problems5
&n Swarm, Evolutionary, and Memetic Computing (SEMCCO 2010), Lecture
Notes in Computer Science, Springer Berlin Heidelberg, Vol. 6466, pp. 583-590.
72. Yang, X. S. (2009) 4Harmony search as a metaheuristic algorithm5
&n Music-inspired harmony search
algorithm, Studies in computational intelligence, Springer Berlin Heidelberg, Vol. 191 pp.1-14.
73. Yang, X. S. (2010) Nature-inspired metaheuristic algorithms. Luniver press.
74. Yang, X. S. (2010). Appendix A: test problems in optimization. Engineering optimization, 261-266.
75. Zhang, J., & Sanderson, A. C. (2009) 4JADE: adaptive differential evolution with optional external
archive5
IEEE Transactions on evolutionary computation, Vol. 13 No. 5, pp.945-958.
76. Zang, H., Zhang, S., & Hapeshi, K. (2010) 4A review of nature-inspired algorithms5
Journal of Bionic
Engineering, Vol. 7 Supplement, pp.S232-S237.
77. Zhang, 
/*


!)$

 
4Differential evolution with dynamic stochastic selection for
constrained optimization5
Information Sciences, Vol. 178 No. 15, pp.3043-3074.





  

  
!$



$%#&%
"%+%!
#!
 * 
 !!
 '#$%*,

 "!#,


#%
 % 
#!
 '#$%*
!
 ,
 ,
!#
!
  # 
#!

'
 '#$%*,
 
 
"!
#!
%
!#
!
 
&%! ,
&-

(!#
$


#$#
!(
!

#!$$.!##
$&""*.
$#&"%!
"#!%
%
%%
!!
!
&$ $$,
 '#$%*

!
 $!#,
 -
&## %*,

$
(!# 
$

 
$$!%
#!$$!#
%
%
*!$$
 $%%&%

!
 !!*,
*!$$
 %# %! 
 '#$%*,
& ,
 -
$
#$#
 %#$%$
 &
!"%+%!

!#%$,
&%.!%'
!"%+%! ,
! % &!&$,
$#%
 
! %!#
!"%+%! ,
&%.
 %
$*$%$,
!")
$*$%$,
!!#%
 % ,
"#!%*
!%'$,
$(#
!"%+%! ,


%!#*,
$.!# + 
$*$%$
 
&%.%!# %
$*$%$-

$
%
!& #
 
#
!
%

"%+%!
 
 %
 !!*
/ 0
$#
-
  
$
"&$
!'#
21
#$#
""#$


"#.#'(
!&# $
 
! # $-

$!
"&$
%(!
#$#
!!$-