artificial bee colony (abc)
DESCRIPTION
TRANSCRIPT
Introduced in 2005 by Dervis Karaboga Honey bee foraging behavior
Types of foraging bee◦ Employed bees
◦ Unemployed bees Scout
Onlooker bees
Picture form www.acclaimclipart.com
Picture form www.acclaimclipart.com , www.computerclipart.com
HiveHive
Dancing area for A
Dancing area for B
Modify from Parmaksızoğlu, S.; Alçı, M. A Novel Cloning Template Designing Method by Using an Artificial Bee Colony Algorithm for Edge Detection of CNN Based Imaging Sensors. Sensors 2011, 11, 5337-5359
Foraging bee
employed bee
Scout
onlookers
Food source initialize(Number of solutions = Employed bees)
Where i = 1,2,…,nn = Food sourcej = Dimension
))(1,0( min,max,min,, jjjji xxrandxx
Send to each solutions (Can be done with Initial phase)◦ Number of solutions = Employed bees
Calculate fitness
)( 1xf)( 2xf
)( 4xf)( 5xf
)( 3xf
Evolve Solution to neighborhood
)( kjijijijij xxxv ijWhere = rand(-1,1)i = 1,2,…,nn = Food sourcej = Dimensionk = rand(1,n)!=i
SolutionXi
j=6
Evolved Solution
Vi
Select better solution
Calculate probability for each solution
Select solution due to probability
Modify from G. Yan et al. “” An Effective Refinement Artificial Bee Colony Optimization AlgorithmBased On Chaotic Search and Application for PID Control Tuning,Journal of Computational Information Systems 7:9 (2011) 3309-3316
n
ii
ii
xf
xfxP
1
)(
)(}{
1Employed bee
Onlooker
2 3 4 5
2 3 4 5 Ri =rand(0,1)1
Ri<P(xi) ?No
Evolve Solution to neighborhood
Select better solution(Same as Employed bee phase)
Modify from G. Yan et al. “” An Effective Refinement Artificial Bee Colony Optimization AlgorithmBased On Chaotic Search and Application for PID Control Tuning,Journal of Computational Information Systems 7:9 (2011) 3309-3316
)( kjijijijij xxxv Where = rand(-1,1)i = 1,2,…,nn = Food sourcej = Dimensionk = rand(1,n)!=i
ij
)(
))(1,0()1( minmaxmin
Gx
xxrandxGx
ii
, counter ≥ limit
, else
No of food source visited = “limit”
Send scouts to find new source
Swarm size Employed bees(50% of swarm) Onlookers(50% of swarm) Scouts(1) Limit Dimension
Modify from D. Karaboga, An idea based on honey bee swarm for numerical optimization. Technical Report TR06, Erciyes University, Engineering Faculty, Computer Engineering Department, 2005
Advantages◦ Few control parameters◦ Fast convergence◦ Both exploration & exploitation
Disadvantages◦ Search space limited by initial solution (normal
distribution sample should use in initialize step)
VEABC◦ Multi-objective ABC◦ Inspired by VEGA, BEPSO
◦ Separate in to k swarm◦ Due to number of objective
◦ Evaluate each swarm to each objective
◦ Position of each swarm -> update neighbor solution
S.N. Omkar, J. Senthilnath, R. Khandelwal, G. Nrayana Naik, S. Gopalakrishnan “Artificial Bee Colony (ABC) for multi-objective design optimization of composite structures
,” Applied Soft Computing, Volume 11, Issue 1, January, 2011
Design composite structure◦ Objectives
Minimize weight Minimize total cost Specified strength
◦ Variables Number of layers Stacking sequence Thickness of each layer
◦ Evaluation Stresses of component Failure criteria
◦ Comparison PSO, AIS, GA
De Jong
M. Molga, C. Smutnicki, “Test functions for optimization needs”,http://www.zsd.ict.pwr.wroc.pl/files/docs/functions.pdf
◦ Function
◦ Decision space
D
iix
1
2
D12.5,12.5 Griewangk
◦ Function
◦ Decision space
1)cos(4000
1
11
2
D
i
iD
ii
i
xx
D600,600
Rastrigin
M. Molga, C. Smutnicki, “Test functions for optimization needs”,http://www.zsd.ict.pwr.wroc.pl/files/docs/functions.pdf
◦ Function
◦ Decision space D12.5,12.5
Rosenbrock
◦ Function
◦ Decision space
D
iii xxD
0
2 ))2cos(10(10
21
0
221 )1()(100 i
D
iii xxx
D048.2,048.2
◦ Swarm size = 10◦ Swarm size = 50◦ Swarm size = 100
0 200 400 600 800 1000 1200 1400 1600 1800 200010
-15
10-10
10-5
100
105
Cycle
Bes
t fu
nctio
n va
lue
De Jong
◦ Swarm size = 10◦ Swarm size = 50◦ Swarm size = 100
Griewangk
0 200 400 600 800 1000 1200 1400 1600 1800 200010
-12
10-10
10-8
10-6
10-4
10-2
100
102
104
Cycle
Bes
t fu
nctio
n va
lue
◦ Swarm size = 10◦ Swarm size = 50◦ Swarm size = 100
Rastrigin
0 500 1000 1500 2000 2500 3000 3500 400010
-15
10-10
10-5
100
105
Cycle
Bes
t fu
nctio
n va
lue
◦ Swarm size = 10◦ Swarm size = 50◦ Swarm size = 100
Rosenbrock
0 500 1000 1500 2000 2500 3000 3500 4000 450010
1
102
103
104
105
Cycle
Bes
t fu
nctio
n va
lue
ne is number of employed bee
Swarm size = 50, D = 50, 30 runs, 5000 evaluations
Function
0.1*D*ne 0.5*D*ne D*ne No scouts
De Jong 1.40E-15 9.97E-16 9.86E-16 1.11E-15
Griewank 2.53E-14 2.52E-16 1.44E-16 0.000350
Rastrigin 5.25E-11 6.07E-16 2.31E-16 0.000336
Rosenbrock 58.310518 58.444032 51.074693 48.912436
Effect of “limit”