SlideShare uma empresa Scribd logo
1 de 8
Baixar para ler offline
Variant of Differential Evolution Algorithm
Richa Shukla, Bramah Hazela, Shashwat Shukla, Ravi Prakash*,
Krishn Kumar Mishra*
Computer Science and Engineering Department
Amity University , Lucknow India, 226028 and
Computer Science and Engineering Department
MNNIT Allahabad, Allahabad India, 211004*
Email: richamity22@gmail.com, bhazela@lko.amity.edu, sshukla1@lko.amity.edu,
raviprakashguddu@gmail.com, kkm@mnnit.ac.in
Abstract. Differential Evolution is a nature inspired optimization tech-
nique. It has been achieved best solutions on large area of test suits.
DE algorithm is efficient in programming and it has broad applicability
in engineering. This paper presents modified mutation vector genera-
tion strategy of basic DE for solving stagnation problem. A new vari-
ant of differential evolution that is DE New has been proposed and the
performance of DE New is tested in COmparing Continuous Optimis-
ers(COCO) framework on 24 benchmark functions and found DE New
has better exploration performance in search space in comparison to
GA, DE- PSO, DE-AUTO on Black-Box Optimization Benchmarking
(BBOB) 2015 devised by COCO.
Keywords DE New, Exploitation, Robot Kidnapping, Differential Evolution,
mean-cost.
1 Introduction
Optimization problem can be solved with the help of different types of algo-
rithms[1][12]. In this scope reputation of Evolutionary Algorithm(EA) is un-
beatable[2][3][4]. DE is an EA that is used for solving many- modal optimization
problem in continuous search space[5][7]. DE algorithm is efficient and based
on natural selection with globally strong optimization technique in continu-
ous search space[8]. DE logic works with the help of DE operators (mutation,
crossover and selection) and associated parameters[9][10][11]. Myths related to
GA and DE creates due to similarity of the operators. In GA, crossover is the
main operator but in DE mutation operator plays leader role for solving global
optimization problem[13][14]. DE is better in compare to GA because it dose not
use binary encoding[15][4] and probability density function[16]. DE variants are
very helpful for enhancing the vector position, resolve the stagnation problem
and increase the convergence speed. DE basically works on global search space.
So for better exploration of solution, here DE New navigates to investigate the
local optimum problem for solving blocked solutions. Further content of this
paper has been ordered as below section 2 defines Differential Evolution, sec-
tion 3 states existing and modified algorithm, section 4 presents proposed work,
practical approach of result analysis has been elaborated in section 5 and finally
conclusions are drawn in section 6.
2 Differential Evolution
Differential Evolution is a randomized, population based, meta-heuristic, nature
inspired[17] and global optimization technique[18] given by Storn and Price in
1995[6]. Mutation, crossover and selection are the DE operators[19] but DE algo-
rithm is fully depend on mutation operator. The role of mutation operator is to
generate a mutant vector and the role of crossover is to develop trial vector and
the aim of selection operator is to select best between trial vector and mutant
vector for each generation.
Best example of DE is that it uses in mobile robot systems[20] for determining
the position and orientations of the robot. The way of initialization of the robot
position and track the position is the concept of Robot kidnapping(particle kid-
napping)[21][22][23]. Robot kidnapping understands the position and movement
of robot path without telling the new position[18].
Brief description of DE operators are given below:
1.Mutation: In this operator generate a mutant vector. Select three vectors
randomly from the search space and multiply the difference of two vectors with
a constant value ’F’(range of F is in between 0-2) and add with third vector that
produces a mutant vector mi,g , this process is called mutation. mutant vector
mi,g for each generation is stated as:
xi,g=xp+F(xq-xr) (1)
Notations:
1. i=1,2,......population size.
2. p, q, r are the three random vectors of the search space for ’g’ genera-
tion[24][27].
2.Crossover: After generating of mutant vector, create trial vector with the
help of mutant operator by recombining of two different vectors.
ti,g+1=t1i,g,t2i,g,t3i,g...........................tni,g (2)
The overall equation of trial vector for crossover operator is divided in two
phases , compare randomly generated number with crossover constant value
If random number generated is ≤ CR then we will use mutant vector mi,g
for each generation otherwise choose target vector without altering parent vec-
tor[25][26][28].
3.Selection: Selection of new member is totally depend on selection operator.
Comparison between trial vector and mutant vector, must select best between
these two vector ti,g and xi,g is the aim of selection operator.
3 Existing and Modified DE Algorithms
Algorithm 1 Basic Differential Evolution Algorithm
1: for Each particle i do
2: for each vectors xi,g of population N do
3: MUTATION
4: select three random vectors from search space for mutation
5: xi,g=xp+F(xq-xr)
6: CROSS-OVER
7: if random number generated is ≤ CR then select mi,g
8: else select target vector
9: end if
10: SELECTION
11: evaluate fitness of xi,g and ti,g+1
12: choose best fitness as solution and discard previous
13: End if termination conditions met
14: end for
15: end for
Algorithm 2 Proposed Differential Evolution Algorithm
1: for Each particle i do
2: for each vectors xi,g of population N do
3: MUTATION
4: select xp, xq, xr
5: Fix first vector xp
6: where xq > mean-cost and xr< mean-cost
7: xi,g=xp+F(xq-xr)
8: CROSS-OVER
9: if random number generated is ≤ CR then
10: select mi,g for each generation
11: else select target vector
12: end if
13: SELECTION
14: evaluate fitness of xi,g and ti,g+1
15: choose best fitness as solution and discard previous
16: End if termination conditions met
17: end for
18: end for
4 Proposed Work
In DE algorithm mutation and associated parameters maintain the performance
of search space. There are many solutions in search space that are blocked,
stagnated, trapped and unable to give better solution for multi-modal prob-
lems in global search space. So for solving this problem DE NEW algorithm has
been proposed. In this algorithm, strategy of basic mutation operator has been
changed and a new variant of DE algorithm (DE NEW) has been proposed in
this paper. The logic of proposed DE NEW algorithm is to select three vectors
from the search space xp, xq, xr. Calculate the mean-cost and on the basis of
mean-cost, search space has been divided into two parts. Select xpvector ran-
domly from the total population size and stored in index number c, it will be the
fixed vector in each iterations. Then select next two vectors xq and xr. Choose
xq randomly from the above mean cost and stored in index number a and select
xr randomly from the below mean cost and stored in index number b.The new
strategy of mutation operator has been used for providing better exploration of
solution and greater diversity of solutions in search space in compare to GA, DE-
PSO, DE-AUTO on Black-Box Optimization Benchmarking (BBOB)[29] 2015
devised by COCO.
Fig. 1: DE NEW
5 Result Analysis
The three popular algorithm’s have been compared with tested performance of
DE New. In below table the rank has been alloted for proposed variant and
noiseless results have been attached. Its performance decreased might be due
to lower value of randomized F, Cr etc. But in this paper, DE New has been
succeeded to give better results for each modal in 24 benchmark functions in
2D, 3D, 5D, 10D, 20D and 40D.
Fig. 2: Comparison Results
Table 1: Ranking of Proposed Variant
DIMENSIONS hcond lcond mult2 multi separable
2D 2nd 2nd 2nd 1st 1st
3D 2nd 2nd 3rd 2nd 1st
5D 2nd 2nd 2nd 2nd 2nd
10D 2nd 2nd 2nd 2nd 2nd
20D 2nd 2nd 2nd 1st 2nd
40D 2nd 2nd 4th 1st 2nd
6 Conclusion
In this paper, proposed variant of DE algorithm has been based on modified mu-
tation strategy. It gives best performance for unimodal and multi-modal prob-
lems and increase the convergence rate and resolve the stagnation problems.
In DE New unlike basic DE and other algorithms, it can get best optimization
performance. DE New is able to give satisfactory performance for DE-AUTO
and best performance for DE-PSO, GA in 2D, 3D, 5D, 10D, 20D and 40D on
various benchmark functions, however there are a lot of ways for improving it.
In future work, DE algorithm can be used in cost estimation of the softwares
and implemented in many applications. Also merge two or more Evolutionary
Algorithm for achieving desired results.
References
[1] Z. W. Geem, J. H. Kim, and G. Loganathan, “A new heuristic optimization
algorithm: harmony search,” Simulation, vol. 76, no. 2, pp. 60–68, 2001.
[2] R. Storn and K. Price, “Differential evolution–a simple and efficient heuristic
for global optimization over continuous spaces,” Journal of global optimization,
vol. 11, no. 4, pp. 341–359, 1997.
[3] S. Tiwari, K. Mishra, and A. K. Misra, “Test case generation for modified code
using a variant of particle swarm optimization (pso) algorithm,” in Information
Technology: New Generations (ITNG), 2013 Tenth International Conference on.
IEEE, 2013, pp. 363–368.
[4] K. Mishra, S. Tiwari, and A. Misra, “Combining non revisiting genetic al-
gorithm and neural network to generate test cases for white box testing,” in
Practical Applications of Intelligent Systems. Springer, 2011, pp. 373–380.
[5] E. Mezura-Montes, J. Vel azquez-Reyes, and C. A. Coello Coello, “A com-
parative study of differential evolution variants for global optimization,” in Pro-
ceedings of the 8th annual conference on Genetic and evolutionary computation.
ACM, 2006, pp. 485–492.
[6] R. Storn and K. Price, Differential evolution-a simple and efficient adaptive
scheme for global optimization over continuous spaces. ICSI Berkeley, 1995, vol.
3.
[7] D. Zaharie, “Critical values for the control parameters of differential evolu-
tion algorithms,” in Proceedings of MENDEL, vol. 2, 2002, p. 6267.
[8] Y. Ao and H. Chi, “Experimental study on differential evolution strategies,”
in Intelligent Systems, 2009. GCIS’09. WRI Global Congress on, vol. 2. IEEE,
2009, pp. 19–24.
[9] M. Daoudi, S. Hamena, Z. Benmounah, and M. Batouche, “Parallel diffrential
evolution clustering algorithm based on mapreduce,” in Soft Computing and Pat-
tern Recognition (SoCPaR), 2014 6th International Conference of. IEEE, 2014,
pp. 337–341.
[10] M. F. Tasgetiren, O. Bulut, Q.-K. Pan, and P. N. Suganthan, “A differen-
tial evolution algorithm for the median cycle problem,” in Differential Evolution
(SDE), 2011 IEEE Symposium on. IEEE, 2011, pp. 1–7.
[11] F. Neri and V. Tirronen, “Recent advances in differential evolution: a survey
and experimental analysis,” Artificial Intelligence Review, vol. 33, no. 1-2, pp.
61–106, 2010.
[12] S. Das and P. N. Suganthan, “Differential evolution: a survey of the stateof-
the-art,” Evolutionary Computation, IEEE Transactions on, vol. 15, no. 1, pp.
4–31, 2011.
[13] R. Joshi and A. C. Sanderson, “Minimal representation multisensor fusion
using differential evolution,” Systems, Man and Cybernetics, Part A: Systems
and Humans, IEEE Transactions on, vol. 29, no. 1, pp. 63– 76, 1999.
[14] J. Holland, “Adaption in natural and artificial systems,” Ann Arbor, MI:
University of Michigan Press, 1975.
[15] D. E. Goldberg et al., Genetic algorithms in search optimization and ma-
chine learning. Addison-wesley Reading Menlo Park, 1989, vol. 412.
[16] H.-P. Schwefel, “Evolution and optimization seeking,”John Wiley Sons,
1995.
[17] X.-S. Yang, Nature-inspired metaheuristic algorithms. Luniver press, 2010.
[18] A. R. Vahdat, N. NourAshrafoddin, and S. S. Ghidary, “Mobile robot global
localization using differential evolution and particle swarm optimization,” in
Evolutionary Computation, 2007. CEC 2007. IEEE Congress on. IEEE, 2007,
pp. 1527–1534.
[19] D. Kumar and K. K. Mishra, “Incorporating logic in artificial bee colony
(abc) algorithm to solve first order logic problems: The logical abc,” in Knowl-
edge and Smart Technology (KST), 2015 7th International Conference on. IEEE,
2015, pp. 65–70.
[20] C. Gonz alez, D. Blanco, and L. Moreno, “Optimum robot manipulator path
generation using differential evolution,” in Evolutionary Computation, 2009.
CEC’09. IEEE Congress on. IEEE, 2009, pp. 3322–3329.
[21] J. Gu and G. Gu, “Differential evolution with a local search operator,” in In-
formatics in Control, Automation and Robotics (CAR), 2010 2nd International
Asia Conference on, vol. 2. IEEE, 2010, pp. 480–483.
[22] J. M. Hereford, M. Siebold, and S. Nichols, “Using the particle swarm opti-
mization algorithm for robotic search applications,” in Swarm Intelligence Sym-
posium, 2007. SIS 2007. IEEE. IEEE, 2007, pp. 53–59.
[23] V. Pano and P. R. Ouyang, “Comparative study of ga, pso, and de for tuning
position domain pid controller,” in Robotics and Biomimetics (ROBIO), 2014
IEEE International Conference on. IEEE, 2014, pp. 1254–1259.
[24] M. G. Epitropakis, D. K. Tasoulis, N. G. Pavlidis, V. P. Plagianakos, and
M. N. Vrahatis, “Enhancing differential evolution utilizing proximitybased mu-
tation operators,” Evolutionary Computation, IEEE Transactions on, vol. 15,
no. 1, pp. 99–119, 2011.
[25] C. Sun, H. Zhou, and L. Chen, “Improved differential evolution algorithms,”
in Computer Science and Automation Engineering (CSAE), 2012 IEEE Interna-
tional Conference on, vol. 3. IEEE, 2012, pp. 142– 145.
[26] A. Tripathi, N. Saxena, K. K. Mishra, and A. K. Misra, “An environmen-
tal adaption method with real parameter encoding for dynamic environment,”
Journal of Intelligent Fuzzy Systems, no. Preprint, pp. 1–13.
[27] P. Melin, F. Olivas, O. Castillo, F. Valdez, J. Soria, and M. Valdez, “Op-
timal design of fuzzy classification systems using pso with dynamic parameter
adaptation through fuzzy logic,” Expert Systems with Applications, vol. 40, no.
8, pp. 3196–3206, 2013.
[28] Y.-W. Chen, C.-L. Lin, and A. Mimori, “Multimodal medical image regis-
tration using particle swarm optimization,” in Intelligent Systems Design and
Applications, 2008. ISDA’08. Eighth International Conference on, vol.3. IEEE,
2008, pp. 127–131.
[29] Brockho. (n.d.). Cec-bbob-2015 [COmparing Continuous Optimisers: COCO].
Retrieved June 11, 2016, from http://coco.gforge.inria.fr/

Mais conteúdo relacionado

Mais procurados

Developing effective meta heuristics for a probabilistic
Developing effective meta heuristics for a probabilisticDeveloping effective meta heuristics for a probabilistic
Developing effective meta heuristics for a probabilistic
Hari Rajagopalan
 
EXPERIMENTS ON HYPOTHESIS "FUZZY K-MEANS IS BETTER THAN K-MEANS FOR CLUSTERING"
EXPERIMENTS ON HYPOTHESIS "FUZZY K-MEANS IS BETTER THAN K-MEANS FOR CLUSTERING"EXPERIMENTS ON HYPOTHESIS "FUZZY K-MEANS IS BETTER THAN K-MEANS FOR CLUSTERING"
EXPERIMENTS ON HYPOTHESIS "FUZZY K-MEANS IS BETTER THAN K-MEANS FOR CLUSTERING"
IJDKP
 

Mais procurados (18)

Post-optimal analysis of LPP
Post-optimal analysis of LPPPost-optimal analysis of LPP
Post-optimal analysis of LPP
 
Universal Portfolios Generated by Reciprocal Functions of Price Relatives
Universal Portfolios Generated by Reciprocal Functions of Price RelativesUniversal Portfolios Generated by Reciprocal Functions of Price Relatives
Universal Portfolios Generated by Reciprocal Functions of Price Relatives
 
KMAP PAPER (1)
KMAP PAPER (1)KMAP PAPER (1)
KMAP PAPER (1)
 
Developing effective meta heuristics for a probabilistic
Developing effective meta heuristics for a probabilisticDeveloping effective meta heuristics for a probabilistic
Developing effective meta heuristics for a probabilistic
 
Multi-Index Bi-Criterion Transportation Problem: A Fuzzy Approach
Multi-Index Bi-Criterion Transportation Problem: A Fuzzy ApproachMulti-Index Bi-Criterion Transportation Problem: A Fuzzy Approach
Multi-Index Bi-Criterion Transportation Problem: A Fuzzy Approach
 
MIXED 0−1 GOAL PROGRAMMING APPROACH TO INTERVAL-VALUED BILEVEL PROGRAMMING PR...
MIXED 0−1 GOAL PROGRAMMING APPROACH TO INTERVAL-VALUED BILEVEL PROGRAMMING PR...MIXED 0−1 GOAL PROGRAMMING APPROACH TO INTERVAL-VALUED BILEVEL PROGRAMMING PR...
MIXED 0−1 GOAL PROGRAMMING APPROACH TO INTERVAL-VALUED BILEVEL PROGRAMMING PR...
 
Duality in Linear Programming Problem
Duality in Linear Programming ProblemDuality in Linear Programming Problem
Duality in Linear Programming Problem
 
Optimization of Mechanical Design Problems Using Improved Differential Evolut...
Optimization of Mechanical Design Problems Using Improved Differential Evolut...Optimization of Mechanical Design Problems Using Improved Differential Evolut...
Optimization of Mechanical Design Problems Using Improved Differential Evolut...
 
Parallel Guided Local Search and Some Preliminary Experimental Results for Co...
Parallel Guided Local Search and Some Preliminary Experimental Results for Co...Parallel Guided Local Search and Some Preliminary Experimental Results for Co...
Parallel Guided Local Search and Some Preliminary Experimental Results for Co...
 
Unit.4.integer programming
Unit.4.integer programmingUnit.4.integer programming
Unit.4.integer programming
 
Unit.3. duality and sensetivity analisis
Unit.3. duality and sensetivity analisisUnit.3. duality and sensetivity analisis
Unit.3. duality and sensetivity analisis
 
EXPERIMENTS ON HYPOTHESIS "FUZZY K-MEANS IS BETTER THAN K-MEANS FOR CLUSTERING"
EXPERIMENTS ON HYPOTHESIS "FUZZY K-MEANS IS BETTER THAN K-MEANS FOR CLUSTERING"EXPERIMENTS ON HYPOTHESIS "FUZZY K-MEANS IS BETTER THAN K-MEANS FOR CLUSTERING"
EXPERIMENTS ON HYPOTHESIS "FUZZY K-MEANS IS BETTER THAN K-MEANS FOR CLUSTERING"
 
True global optimality of the pressure vessel design problem: a benchmark for...
True global optimality of the pressure vessel design problem: a benchmark for...True global optimality of the pressure vessel design problem: a benchmark for...
True global optimality of the pressure vessel design problem: a benchmark for...
 
Duality in nonlinear fractional programming problem using fuzzy programming a...
Duality in nonlinear fractional programming problem using fuzzy programming a...Duality in nonlinear fractional programming problem using fuzzy programming a...
Duality in nonlinear fractional programming problem using fuzzy programming a...
 
352735343 rsh-qam11-tif-07-doc
352735343 rsh-qam11-tif-07-doc352735343 rsh-qam11-tif-07-doc
352735343 rsh-qam11-tif-07-doc
 
SOLVING OPTIMAL COMPONENTS ASSIGNMENT PROBLEM FOR A MULTISTATE NETWORK USING ...
SOLVING OPTIMAL COMPONENTS ASSIGNMENT PROBLEM FOR A MULTISTATE NETWORK USING ...SOLVING OPTIMAL COMPONENTS ASSIGNMENT PROBLEM FOR A MULTISTATE NETWORK USING ...
SOLVING OPTIMAL COMPONENTS ASSIGNMENT PROBLEM FOR A MULTISTATE NETWORK USING ...
 
SOLVING OPTIMAL COMPONENTS ASSIGNMENT PROBLEM FOR A MULTISTATE NETWORK USING ...
SOLVING OPTIMAL COMPONENTS ASSIGNMENT PROBLEM FOR A MULTISTATE NETWORK USING ...SOLVING OPTIMAL COMPONENTS ASSIGNMENT PROBLEM FOR A MULTISTATE NETWORK USING ...
SOLVING OPTIMAL COMPONENTS ASSIGNMENT PROBLEM FOR A MULTISTATE NETWORK USING ...
 
A h k clustering algorithm for high dimensional data using ensemble learning
A h k clustering algorithm for high dimensional data using ensemble learningA h k clustering algorithm for high dimensional data using ensemble learning
A h k clustering algorithm for high dimensional data using ensemble learning
 

Destaque

Sabahattin ali kagni
Sabahattin ali kagniSabahattin ali kagni
Sabahattin ali kagni
barstugan
 
Sinerxias entre o sector turístico e o audiovisual
Sinerxias entre o sector turístico e o audiovisualSinerxias entre o sector turístico e o audiovisual
Sinerxias entre o sector turístico e o audiovisual
Lucia Costas
 
Grupna kupovina Beograd Srbija, Snizenje Beograd
Grupna kupovina Beograd Srbija, Snizenje BeogradGrupna kupovina Beograd Srbija, Snizenje Beograd
Grupna kupovina Beograd Srbija, Snizenje Beograd
Grupna kupovina
 
Collective intelligence in web 2
Collective intelligence in web 2Collective intelligence in web 2
Collective intelligence in web 2
matty91
 
Mano miestas mycka
Mano miestas myckaMano miestas mycka
Mano miestas mycka
Mickute
 

Destaque (18)

Sabahattin ali kagni
Sabahattin ali kagniSabahattin ali kagni
Sabahattin ali kagni
 
SMX East 2012 - Managing Management Effectively - PPC
SMX East 2012 - Managing Management Effectively - PPCSMX East 2012 - Managing Management Effectively - PPC
SMX East 2012 - Managing Management Effectively - PPC
 
Tracce di fede
Tracce di fedeTracce di fede
Tracce di fede
 
TRiBE! collective monkey poet sample
TRiBE! collective monkey poet sampleTRiBE! collective monkey poet sample
TRiBE! collective monkey poet sample
 
Sinerxias entre o sector turístico e o audiovisual
Sinerxias entre o sector turístico e o audiovisualSinerxias entre o sector turístico e o audiovisual
Sinerxias entre o sector turístico e o audiovisual
 
το αύριο και οι φόβοι μας
το αύριο και οι φόβοι μαςτο αύριο και οι φόβοι μας
το αύριο και οι φόβοι μας
 
Grupna kupovina Beograd Srbija, Snizenje Beograd
Grupna kupovina Beograd Srbija, Snizenje BeogradGrupna kupovina Beograd Srbija, Snizenje Beograd
Grupna kupovina Beograd Srbija, Snizenje Beograd
 
Iot and-gnu-linux
Iot and-gnu-linuxIot and-gnu-linux
Iot and-gnu-linux
 
Venta spanish
Venta spanishVenta spanish
Venta spanish
 
Интерактивный урок "Введение в алгебру"
Интерактивный урок "Введение в алгебру"Интерактивный урок "Введение в алгебру"
Интерактивный урок "Введение в алгебру"
 
Ciudad de Cuenca
Ciudad de CuencaCiudad de Cuenca
Ciudad de Cuenca
 
Collective intelligence in web 2
Collective intelligence in web 2Collective intelligence in web 2
Collective intelligence in web 2
 
IAM isnt Magic
IAM isnt MagicIAM isnt Magic
IAM isnt Magic
 
Making waves
Making wavesMaking waves
Making waves
 
Mano miestas mycka
Mano miestas myckaMano miestas mycka
Mano miestas mycka
 
6197 Son Hali
6197 Son Hali6197 Son Hali
6197 Son Hali
 
Tugasan 4
Tugasan 4Tugasan 4
Tugasan 4
 
Terapia cognitiva conductual
Terapia cognitiva conductualTerapia cognitiva conductual
Terapia cognitiva conductual
 

Semelhante a Updated paper in latex(1)

Mimo system-order-reduction-using-real-coded-genetic-algorithm
Mimo system-order-reduction-using-real-coded-genetic-algorithmMimo system-order-reduction-using-real-coded-genetic-algorithm
Mimo system-order-reduction-using-real-coded-genetic-algorithm
Cemal Ardil
 
Performance Analysis of Genetic Algorithm as a Stochastic Optimization Tool i...
Performance Analysis of Genetic Algorithm as a Stochastic Optimization Tool i...Performance Analysis of Genetic Algorithm as a Stochastic Optimization Tool i...
Performance Analysis of Genetic Algorithm as a Stochastic Optimization Tool i...
paperpublications3
 

Semelhante a Updated paper in latex(1) (20)

Optimization of Mechanical Design Problems Using Improved Differential Evolut...
Optimization of Mechanical Design Problems Using Improved Differential Evolut...Optimization of Mechanical Design Problems Using Improved Differential Evolut...
Optimization of Mechanical Design Problems Using Improved Differential Evolut...
 
Medical diagnosis classification
Medical diagnosis classificationMedical diagnosis classification
Medical diagnosis classification
 
MEDICAL DIAGNOSIS CLASSIFICATION USING MIGRATION BASED DIFFERENTIAL EVOLUTION...
MEDICAL DIAGNOSIS CLASSIFICATION USING MIGRATION BASED DIFFERENTIAL EVOLUTION...MEDICAL DIAGNOSIS CLASSIFICATION USING MIGRATION BASED DIFFERENTIAL EVOLUTION...
MEDICAL DIAGNOSIS CLASSIFICATION USING MIGRATION BASED DIFFERENTIAL EVOLUTION...
 
Two-Stage Eagle Strategy with Differential Evolution
Two-Stage Eagle Strategy with Differential EvolutionTwo-Stage Eagle Strategy with Differential Evolution
Two-Stage Eagle Strategy with Differential Evolution
 
Newtonian Law Inspired Optimization Techniques Based on Gravitational Search ...
Newtonian Law Inspired Optimization Techniques Based on Gravitational Search ...Newtonian Law Inspired Optimization Techniques Based on Gravitational Search ...
Newtonian Law Inspired Optimization Techniques Based on Gravitational Search ...
 
Interactive Fuzzy Goal Programming approach for Tri-Level Linear Programming ...
Interactive Fuzzy Goal Programming approach for Tri-Level Linear Programming ...Interactive Fuzzy Goal Programming approach for Tri-Level Linear Programming ...
Interactive Fuzzy Goal Programming approach for Tri-Level Linear Programming ...
 
An optimal design of current conveyors using a hybrid-based metaheuristic alg...
An optimal design of current conveyors using a hybrid-based metaheuristic alg...An optimal design of current conveyors using a hybrid-based metaheuristic alg...
An optimal design of current conveyors using a hybrid-based metaheuristic alg...
 
Ds33717725
Ds33717725Ds33717725
Ds33717725
 
Ds33717725
Ds33717725Ds33717725
Ds33717725
 
A Real Coded Genetic Algorithm For Solving Integer And Mixed Integer Optimiza...
A Real Coded Genetic Algorithm For Solving Integer And Mixed Integer Optimiza...A Real Coded Genetic Algorithm For Solving Integer And Mixed Integer Optimiza...
A Real Coded Genetic Algorithm For Solving Integer And Mixed Integer Optimiza...
 
Mimo system-order-reduction-using-real-coded-genetic-algorithm
Mimo system-order-reduction-using-real-coded-genetic-algorithmMimo system-order-reduction-using-real-coded-genetic-algorithm
Mimo system-order-reduction-using-real-coded-genetic-algorithm
 
Using particle swarm optimization to solve test functions problems
Using particle swarm optimization to solve test functions problemsUsing particle swarm optimization to solve test functions problems
Using particle swarm optimization to solve test functions problems
 
MULTI-OBJECTIVE ENERGY EFFICIENT OPTIMIZATION ALGORITHM FOR COVERAGE CONTROL ...
MULTI-OBJECTIVE ENERGY EFFICIENT OPTIMIZATION ALGORITHM FOR COVERAGE CONTROL ...MULTI-OBJECTIVE ENERGY EFFICIENT OPTIMIZATION ALGORITHM FOR COVERAGE CONTROL ...
MULTI-OBJECTIVE ENERGY EFFICIENT OPTIMIZATION ALGORITHM FOR COVERAGE CONTROL ...
 
A Mathematical Programming Approach for Selection of Variables in Cluster Ana...
A Mathematical Programming Approach for Selection of Variables in Cluster Ana...A Mathematical Programming Approach for Selection of Variables in Cluster Ana...
A Mathematical Programming Approach for Selection of Variables in Cluster Ana...
 
Fulltext
FulltextFulltext
Fulltext
 
Performance Analysis of Genetic Algorithm as a Stochastic Optimization Tool i...
Performance Analysis of Genetic Algorithm as a Stochastic Optimization Tool i...Performance Analysis of Genetic Algorithm as a Stochastic Optimization Tool i...
Performance Analysis of Genetic Algorithm as a Stochastic Optimization Tool i...
 
A HYBRID COA/ε-CONSTRAINT METHOD FOR SOLVING MULTI-OBJECTIVE PROBLEMS
A HYBRID COA/ε-CONSTRAINT METHOD FOR SOLVING MULTI-OBJECTIVE PROBLEMSA HYBRID COA/ε-CONSTRAINT METHOD FOR SOLVING MULTI-OBJECTIVE PROBLEMS
A HYBRID COA/ε-CONSTRAINT METHOD FOR SOLVING MULTI-OBJECTIVE PROBLEMS
 
Hybrid method for achieving Pareto front on economic emission dispatch
Hybrid method for achieving Pareto front on economic  emission dispatch Hybrid method for achieving Pareto front on economic  emission dispatch
Hybrid method for achieving Pareto front on economic emission dispatch
 
A HYBRID COA/ε-CONSTRAINT METHOD FOR SOLVING MULTI-OBJECTIVE PROBLEMS
A HYBRID COA/ε-CONSTRAINT METHOD FOR SOLVING MULTI-OBJECTIVE PROBLEMSA HYBRID COA/ε-CONSTRAINT METHOD FOR SOLVING MULTI-OBJECTIVE PROBLEMS
A HYBRID COA/ε-CONSTRAINT METHOD FOR SOLVING MULTI-OBJECTIVE PROBLEMS
 
Energy management system
Energy management systemEnergy management system
Energy management system
 

Updated paper in latex(1)

  • 1. Variant of Differential Evolution Algorithm Richa Shukla, Bramah Hazela, Shashwat Shukla, Ravi Prakash*, Krishn Kumar Mishra* Computer Science and Engineering Department Amity University , Lucknow India, 226028 and Computer Science and Engineering Department MNNIT Allahabad, Allahabad India, 211004* Email: richamity22@gmail.com, bhazela@lko.amity.edu, sshukla1@lko.amity.edu, raviprakashguddu@gmail.com, kkm@mnnit.ac.in Abstract. Differential Evolution is a nature inspired optimization tech- nique. It has been achieved best solutions on large area of test suits. DE algorithm is efficient in programming and it has broad applicability in engineering. This paper presents modified mutation vector genera- tion strategy of basic DE for solving stagnation problem. A new vari- ant of differential evolution that is DE New has been proposed and the performance of DE New is tested in COmparing Continuous Optimis- ers(COCO) framework on 24 benchmark functions and found DE New has better exploration performance in search space in comparison to GA, DE- PSO, DE-AUTO on Black-Box Optimization Benchmarking (BBOB) 2015 devised by COCO. Keywords DE New, Exploitation, Robot Kidnapping, Differential Evolution, mean-cost. 1 Introduction Optimization problem can be solved with the help of different types of algo- rithms[1][12]. In this scope reputation of Evolutionary Algorithm(EA) is un- beatable[2][3][4]. DE is an EA that is used for solving many- modal optimization problem in continuous search space[5][7]. DE algorithm is efficient and based on natural selection with globally strong optimization technique in continu- ous search space[8]. DE logic works with the help of DE operators (mutation, crossover and selection) and associated parameters[9][10][11]. Myths related to GA and DE creates due to similarity of the operators. In GA, crossover is the main operator but in DE mutation operator plays leader role for solving global optimization problem[13][14]. DE is better in compare to GA because it dose not use binary encoding[15][4] and probability density function[16]. DE variants are very helpful for enhancing the vector position, resolve the stagnation problem and increase the convergence speed. DE basically works on global search space. So for better exploration of solution, here DE New navigates to investigate the local optimum problem for solving blocked solutions. Further content of this
  • 2. paper has been ordered as below section 2 defines Differential Evolution, sec- tion 3 states existing and modified algorithm, section 4 presents proposed work, practical approach of result analysis has been elaborated in section 5 and finally conclusions are drawn in section 6. 2 Differential Evolution Differential Evolution is a randomized, population based, meta-heuristic, nature inspired[17] and global optimization technique[18] given by Storn and Price in 1995[6]. Mutation, crossover and selection are the DE operators[19] but DE algo- rithm is fully depend on mutation operator. The role of mutation operator is to generate a mutant vector and the role of crossover is to develop trial vector and the aim of selection operator is to select best between trial vector and mutant vector for each generation. Best example of DE is that it uses in mobile robot systems[20] for determining the position and orientations of the robot. The way of initialization of the robot position and track the position is the concept of Robot kidnapping(particle kid- napping)[21][22][23]. Robot kidnapping understands the position and movement of robot path without telling the new position[18]. Brief description of DE operators are given below: 1.Mutation: In this operator generate a mutant vector. Select three vectors randomly from the search space and multiply the difference of two vectors with a constant value ’F’(range of F is in between 0-2) and add with third vector that produces a mutant vector mi,g , this process is called mutation. mutant vector mi,g for each generation is stated as: xi,g=xp+F(xq-xr) (1) Notations: 1. i=1,2,......population size. 2. p, q, r are the three random vectors of the search space for ’g’ genera- tion[24][27]. 2.Crossover: After generating of mutant vector, create trial vector with the help of mutant operator by recombining of two different vectors. ti,g+1=t1i,g,t2i,g,t3i,g...........................tni,g (2) The overall equation of trial vector for crossover operator is divided in two phases , compare randomly generated number with crossover constant value If random number generated is ≤ CR then we will use mutant vector mi,g for each generation otherwise choose target vector without altering parent vec- tor[25][26][28]. 3.Selection: Selection of new member is totally depend on selection operator. Comparison between trial vector and mutant vector, must select best between these two vector ti,g and xi,g is the aim of selection operator.
  • 3. 3 Existing and Modified DE Algorithms Algorithm 1 Basic Differential Evolution Algorithm 1: for Each particle i do 2: for each vectors xi,g of population N do 3: MUTATION 4: select three random vectors from search space for mutation 5: xi,g=xp+F(xq-xr) 6: CROSS-OVER 7: if random number generated is ≤ CR then select mi,g 8: else select target vector 9: end if 10: SELECTION 11: evaluate fitness of xi,g and ti,g+1 12: choose best fitness as solution and discard previous 13: End if termination conditions met 14: end for 15: end for Algorithm 2 Proposed Differential Evolution Algorithm 1: for Each particle i do 2: for each vectors xi,g of population N do 3: MUTATION 4: select xp, xq, xr 5: Fix first vector xp 6: where xq > mean-cost and xr< mean-cost 7: xi,g=xp+F(xq-xr) 8: CROSS-OVER 9: if random number generated is ≤ CR then 10: select mi,g for each generation 11: else select target vector 12: end if 13: SELECTION 14: evaluate fitness of xi,g and ti,g+1 15: choose best fitness as solution and discard previous 16: End if termination conditions met 17: end for 18: end for 4 Proposed Work In DE algorithm mutation and associated parameters maintain the performance of search space. There are many solutions in search space that are blocked,
  • 4. stagnated, trapped and unable to give better solution for multi-modal prob- lems in global search space. So for solving this problem DE NEW algorithm has been proposed. In this algorithm, strategy of basic mutation operator has been changed and a new variant of DE algorithm (DE NEW) has been proposed in this paper. The logic of proposed DE NEW algorithm is to select three vectors from the search space xp, xq, xr. Calculate the mean-cost and on the basis of mean-cost, search space has been divided into two parts. Select xpvector ran- domly from the total population size and stored in index number c, it will be the fixed vector in each iterations. Then select next two vectors xq and xr. Choose xq randomly from the above mean cost and stored in index number a and select xr randomly from the below mean cost and stored in index number b.The new strategy of mutation operator has been used for providing better exploration of solution and greater diversity of solutions in search space in compare to GA, DE- PSO, DE-AUTO on Black-Box Optimization Benchmarking (BBOB)[29] 2015 devised by COCO. Fig. 1: DE NEW 5 Result Analysis The three popular algorithm’s have been compared with tested performance of DE New. In below table the rank has been alloted for proposed variant and noiseless results have been attached. Its performance decreased might be due to lower value of randomized F, Cr etc. But in this paper, DE New has been succeeded to give better results for each modal in 24 benchmark functions in 2D, 3D, 5D, 10D, 20D and 40D.
  • 6. Table 1: Ranking of Proposed Variant DIMENSIONS hcond lcond mult2 multi separable 2D 2nd 2nd 2nd 1st 1st 3D 2nd 2nd 3rd 2nd 1st 5D 2nd 2nd 2nd 2nd 2nd 10D 2nd 2nd 2nd 2nd 2nd 20D 2nd 2nd 2nd 1st 2nd 40D 2nd 2nd 4th 1st 2nd 6 Conclusion In this paper, proposed variant of DE algorithm has been based on modified mu- tation strategy. It gives best performance for unimodal and multi-modal prob- lems and increase the convergence rate and resolve the stagnation problems. In DE New unlike basic DE and other algorithms, it can get best optimization performance. DE New is able to give satisfactory performance for DE-AUTO and best performance for DE-PSO, GA in 2D, 3D, 5D, 10D, 20D and 40D on various benchmark functions, however there are a lot of ways for improving it. In future work, DE algorithm can be used in cost estimation of the softwares and implemented in many applications. Also merge two or more Evolutionary Algorithm for achieving desired results. References [1] Z. W. Geem, J. H. Kim, and G. Loganathan, “A new heuristic optimization algorithm: harmony search,” Simulation, vol. 76, no. 2, pp. 60–68, 2001. [2] R. Storn and K. Price, “Differential evolution–a simple and efficient heuristic for global optimization over continuous spaces,” Journal of global optimization, vol. 11, no. 4, pp. 341–359, 1997. [3] S. Tiwari, K. Mishra, and A. K. Misra, “Test case generation for modified code using a variant of particle swarm optimization (pso) algorithm,” in Information Technology: New Generations (ITNG), 2013 Tenth International Conference on. IEEE, 2013, pp. 363–368. [4] K. Mishra, S. Tiwari, and A. Misra, “Combining non revisiting genetic al- gorithm and neural network to generate test cases for white box testing,” in Practical Applications of Intelligent Systems. Springer, 2011, pp. 373–380. [5] E. Mezura-Montes, J. Vel azquez-Reyes, and C. A. Coello Coello, “A com- parative study of differential evolution variants for global optimization,” in Pro- ceedings of the 8th annual conference on Genetic and evolutionary computation. ACM, 2006, pp. 485–492. [6] R. Storn and K. Price, Differential evolution-a simple and efficient adaptive scheme for global optimization over continuous spaces. ICSI Berkeley, 1995, vol. 3.
  • 7. [7] D. Zaharie, “Critical values for the control parameters of differential evolu- tion algorithms,” in Proceedings of MENDEL, vol. 2, 2002, p. 6267. [8] Y. Ao and H. Chi, “Experimental study on differential evolution strategies,” in Intelligent Systems, 2009. GCIS’09. WRI Global Congress on, vol. 2. IEEE, 2009, pp. 19–24. [9] M. Daoudi, S. Hamena, Z. Benmounah, and M. Batouche, “Parallel diffrential evolution clustering algorithm based on mapreduce,” in Soft Computing and Pat- tern Recognition (SoCPaR), 2014 6th International Conference of. IEEE, 2014, pp. 337–341. [10] M. F. Tasgetiren, O. Bulut, Q.-K. Pan, and P. N. Suganthan, “A differen- tial evolution algorithm for the median cycle problem,” in Differential Evolution (SDE), 2011 IEEE Symposium on. IEEE, 2011, pp. 1–7. [11] F. Neri and V. Tirronen, “Recent advances in differential evolution: a survey and experimental analysis,” Artificial Intelligence Review, vol. 33, no. 1-2, pp. 61–106, 2010. [12] S. Das and P. N. Suganthan, “Differential evolution: a survey of the stateof- the-art,” Evolutionary Computation, IEEE Transactions on, vol. 15, no. 1, pp. 4–31, 2011. [13] R. Joshi and A. C. Sanderson, “Minimal representation multisensor fusion using differential evolution,” Systems, Man and Cybernetics, Part A: Systems and Humans, IEEE Transactions on, vol. 29, no. 1, pp. 63– 76, 1999. [14] J. Holland, “Adaption in natural and artificial systems,” Ann Arbor, MI: University of Michigan Press, 1975. [15] D. E. Goldberg et al., Genetic algorithms in search optimization and ma- chine learning. Addison-wesley Reading Menlo Park, 1989, vol. 412. [16] H.-P. Schwefel, “Evolution and optimization seeking,”John Wiley Sons, 1995. [17] X.-S. Yang, Nature-inspired metaheuristic algorithms. Luniver press, 2010. [18] A. R. Vahdat, N. NourAshrafoddin, and S. S. Ghidary, “Mobile robot global localization using differential evolution and particle swarm optimization,” in Evolutionary Computation, 2007. CEC 2007. IEEE Congress on. IEEE, 2007, pp. 1527–1534. [19] D. Kumar and K. K. Mishra, “Incorporating logic in artificial bee colony (abc) algorithm to solve first order logic problems: The logical abc,” in Knowl- edge and Smart Technology (KST), 2015 7th International Conference on. IEEE, 2015, pp. 65–70. [20] C. Gonz alez, D. Blanco, and L. Moreno, “Optimum robot manipulator path generation using differential evolution,” in Evolutionary Computation, 2009. CEC’09. IEEE Congress on. IEEE, 2009, pp. 3322–3329. [21] J. Gu and G. Gu, “Differential evolution with a local search operator,” in In- formatics in Control, Automation and Robotics (CAR), 2010 2nd International Asia Conference on, vol. 2. IEEE, 2010, pp. 480–483. [22] J. M. Hereford, M. Siebold, and S. Nichols, “Using the particle swarm opti- mization algorithm for robotic search applications,” in Swarm Intelligence Sym- posium, 2007. SIS 2007. IEEE. IEEE, 2007, pp. 53–59.
  • 8. [23] V. Pano and P. R. Ouyang, “Comparative study of ga, pso, and de for tuning position domain pid controller,” in Robotics and Biomimetics (ROBIO), 2014 IEEE International Conference on. IEEE, 2014, pp. 1254–1259. [24] M. G. Epitropakis, D. K. Tasoulis, N. G. Pavlidis, V. P. Plagianakos, and M. N. Vrahatis, “Enhancing differential evolution utilizing proximitybased mu- tation operators,” Evolutionary Computation, IEEE Transactions on, vol. 15, no. 1, pp. 99–119, 2011. [25] C. Sun, H. Zhou, and L. Chen, “Improved differential evolution algorithms,” in Computer Science and Automation Engineering (CSAE), 2012 IEEE Interna- tional Conference on, vol. 3. IEEE, 2012, pp. 142– 145. [26] A. Tripathi, N. Saxena, K. K. Mishra, and A. K. Misra, “An environmen- tal adaption method with real parameter encoding for dynamic environment,” Journal of Intelligent Fuzzy Systems, no. Preprint, pp. 1–13. [27] P. Melin, F. Olivas, O. Castillo, F. Valdez, J. Soria, and M. Valdez, “Op- timal design of fuzzy classification systems using pso with dynamic parameter adaptation through fuzzy logic,” Expert Systems with Applications, vol. 40, no. 8, pp. 3196–3206, 2013. [28] Y.-W. Chen, C.-L. Lin, and A. Mimori, “Multimodal medical image regis- tration using particle swarm optimization,” in Intelligent Systems Design and Applications, 2008. ISDA’08. Eighth International Conference on, vol.3. IEEE, 2008, pp. 127–131. [29] Brockho. (n.d.). Cec-bbob-2015 [COmparing Continuous Optimisers: COCO]. Retrieved June 11, 2016, from http://coco.gforge.inria.fr/