+ All Categories
Home > Documents > Particle Swam CA

Particle Swam CA

Date post: 07-Apr-2018
Category:
Upload: nguyen-anh
View: 224 times
Download: 0 times
Share this document with a friend
29
Particle Swarm optimisation 2002-04- 24 Maurice.Clerc@WriteM e.com Particle Swarm optimisation
Transcript

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 1/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

Particle Swarm optimisation

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 2/29

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 3/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

The ³inventors´ (2)

JamesKennedy

Kennedy_ [email protected]

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 4/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

Cooperation example

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 5/29

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 6/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

The basic idea II

The particles in the swarm co-o perate. Theyexchange information about what they¶ve discovered

in the places they have visited The co-operation need only be very simple. In basic

PSO (which is pretty good!) it is like this: ±  A particle has a neighbourhood associated with it.

 ±  A particle knows the fitnesses of those in its

neighbourhood, and uses the  position of the one with bestfitness.

 ± This position is simply used to adjust the particle¶s velocity

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 7/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

Initialization. Positions

and velocities

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 8/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

The basic idea

Each particle is searching for the optimum

Each particle is moving (can¶t search

otherwise!), and hence has a velocity . Each particle remembers the position it was

in where it had its best result so far (its personal best )

But this would not be much good on its own;  particlesneed hel  p in fig uring out where to search.

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 9/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

The basic idea II

The particles in the swarm co-o perate. Theyexchange information about what they¶ve discovered

in the places they have visited The co-operation need only be very simple. In basic

PSO (which is pretty good!) it is like this: ±  A particle has a neighbourhood associated with it.

 ±  A particle knows the fitnesses of those in its

neighbourhood, and uses the  position of the one with bestfitness.

 ± This position is simply used to adjust the particle¶s velocity

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 10/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

What a particle does

In each timestep, a particle has to move to anew position. It does this by adjusting itsvelocity . ± The adj ustment is essentially this:

 ± The c urrent velocity PLUS 

 ±  A weighted random  portion in the direction of its  personal best PLUS 

 ±  A weighted random  portion in the direction of the

neighbourhood best. H aving worked out a new velocity, its  position is

sim ply its old  position  pl us the new velocity.

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 11/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

Neighbourhoods

geographical

social

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 12/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

Neighbourhoods

Global

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 13/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

The circular 

neighbourhood

Virtual circle

1

5

7

64

3

8 2Particle 1¶s 3-

neighbourhood

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 14/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

Particles Adjust their positions according to a``Psychosocial compromise¶¶ between what an

individual is comfortable with, and what society reckons

Here I

am!

The best

perf. of my

neighbours

My best

perf.

 x p g 

 pi

v

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 15/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

Pseudocodehttp://www.swarmintelligence.org/tutorials.php

Equation (a)

v[] = c0 *v[]+ c1 * rand() * (pbest[] - present[])+ c2 * rand() * (gbest[] - present[])

(in the original method, c0=1, but manyresearchers now play with this parameter)

Equation (b)present[] = present[] + v[]

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 16/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

Pseudocodehttp://www.swarmintelligence.org/tutorials.php

For each particleInitialize particle

END

Do

For each particleCalculate fitness valueIf the fitness value is better than its peronal bestset current value as the new pBest

End

Choose the particle with the best fitness value of all as gBestFor each particle

Calculate particle velocity according equation (a)Update particle position according equation (b)

EndWhile maximum iterations or minimum error criteria is not attained

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 17/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

Pseudocodehttp://www.swarmintelligence.org/tutorials.php

Particles' velocities on each dimensionare clamped to a maximum velocity

Vmax. If the sum of accelerations wouldcause the velocity on that dimension toexceed Vmax, which is a parameterspecified by the user. Then the velocity

on that dimension is limited to Vmax.

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 18/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

The basic algorithm:

again

±

±

°

±±

¯

®

!

t  x prand 

t  x prand 

t v

t v

d d  g 

d d i

,2

,1

,0

,0

1

N  F

N  F

E

f or each particle

update

the

velocity

1)1( t vt  xt  xthen move

f or each component d 

At each time step t 

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 19/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

 Animated illustration

Global

optimum

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 20/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

Parameters

Number of particles

C1 (importance of personal best)

C2 (importance of neighbourhood best)

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 21/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

How to choose

parameters

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 22/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

Parameters

Number of particles

(10²50) are reported as usually

sufficient. C1 (importance of personal best)

C2 (importance of neighbourhood best)

Usually C1+C2 = 4. No good reason other 

than empiricism Vmax ± too low, too slow; too high, too

unstable.

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 23/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

Some functions ...

Rosenbrock 

Griewank  Rastrigin

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 24/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

... and some results

30D function PSO Type 1" Evolutionary

algo.(Angeline 98)

rie ank [±300] 0.003944 0.4033

R astrigin [±5] 82.95618 46.4689

R osenbrock [±10] 50.193877 1610.359

Optimum=0, dimension=30Best result after 40 000 evaluations

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 25/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

Some variants

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 26/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

 Adaptive swarm size

There has been enough

improvement

 but there has been not enough

improvement

although I'm the worst

I'm the best

I try to kill myself 

I try to generate a

new particle

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 27/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

 Adaptive coefficients

The better I

am, the more I

f ollow my ownway

The better is my best

neighbour , the more

I tend to go towardshim

Evrand (0«b)( p- x)

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 28/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

Is this Just A Variation on

Evolutionary Algorithms?

8/6/2019 Particle Swam CA

http://slidepdf.com/reader/full/particle-swam-ca 29/29

Particle Swarm optimisation2002-04-

24

Maurice.Clerc@WriteM

e.com

 An applet

Next « Cellular Automata


Recommended