LISTSERV mailing list manager LISTSERV 16.0

Help for ECJ-INTEREST-L Archives


ECJ-INTEREST-L Archives

ECJ-INTEREST-L Archives


ECJ-INTEREST-L@LISTSERV.GMU.EDU


View:

Message:

[

First

|

Previous

|

Next

|

Last

]

By Topic:

[

First

|

Previous

|

Next

|

Last

]

By Author:

[

First

|

Previous

|

Next

|

Last

]

Font:

Proportional Font

LISTSERV Archives

LISTSERV Archives

ECJ-INTEREST-L Home

ECJ-INTEREST-L Home

ECJ-INTEREST-L  November 2008

ECJ-INTEREST-L November 2008

Subject:

Re: State of PSO in ECJ?

From:

Michael Hart <[log in to unmask]>

Reply-To:

ECJ Evolutionary Computation Toolkit <[log in to unmask]>

Date:

Sat, 22 Nov 2008 15:19:10 +1100

Content-Type:

text/plain

Parts/Attachments:

Parts/Attachments

text/plain (232 lines)

Great Sean, thanks for the response.

I'll keep tweaking and perhaps look at assignNeighborhoodBests.

On 22/11/2008, at 6:08 AM, Sean Luke wrote:

> Michael Hart wrote:
>
>> A couple of questions about PSO in ECJ:
>> - PSOBreeder is hardcoded to only look at the first subpopulation.  
>> Is fixing this just a matter of changing breedPopulation to loop  
>> through all subpopulations?
>
> There shouldn't be any issue.  I'm attaching a version of PSOBreeder  
> with the appropriate modification.  Note that it still only uses one  
> *thread* though.  :-(
>
>> - I haven't been getting particularly good results with PSO in ECJ  
>> and while I'm happy to continue to try and tweak parameters, I was  
>> just wondering how similar ECJ's PSO implementation is to those  
>> described here (http://en.wikipedia.org/wiki/Particle_swarm_optimization 
>> ) and here (http://www.particleswarm.info/standard_pso_2007.c)?  
>> I've tried to figure out which parameters in ECJ correspond to  
>> those in the referenced literature, and whether the same  
>> calculations are being performed, but
>
> ECJ's PSO implementation is new and relatively untested: bt yes,  
> it's basically just like the one shown in Wikipedia.  The  
> neighborhood is defined as a ring made out of the subpopulation  
> array itself, which is fairly common.  You can change the  
> neighborhood by overriding the assignNeighborhoodBests function.   
> One reason it may not converge well is because some versions of PSO  
> have no neighborhood information (which converges faster but more  
> easily gets trapped in local optima).
>
>
> > I'm getting a bit lost in the way the PSOBreeder is breeding. Does  
> the
> > "Probability threshold" p, as specified in standard_pso_2007.c,  
> exist
> > in ECJ's implementation?
>
> I believe 'p', in the c code provided, is a mechanism for allowing  
> arbitrary neighborhood topologies.  It also looks O(n), yeesh.   
> Anyway, ECJ only has one built-in topology and so doesn't have 'p'.
>
> BTW: I am wary of specifications defined by code.  We went with the  
> equations in Kennedy's book and other papers.  There's been some PSO  
> improvements in the field since then: we would definitely welcome  
> improvements in ECJ as well there.
>
> PSO in ECJ is really pretty simple.  The ugly part stems from when  
> you need to write a population out to a stream.  In PSO, the  
> population arrays are not just of individuals, but of all the  
> auxillary statistics those individuals store (previous locations,  
> personal bests, etc.).  We could have stored those in the  
> individuals themselves but elected to do it in population arrays,  
> mostly to allow us to just use standard fitness and  
> DoubleVectorIndividual classes.  That results in a lot of I/O code  
> in the PSOSubpopulation class.  Don't let that spook you, it's all  
> boilerplate.
>
> Sean
> /*
>  Copyright 2006 by Ankur Desai, Sean Luke, and George Mason University
>  Licensed under the Academic Free License version 3.0
>  See the file "LICENSE" for more information
> */
>
> package ec.pso;
>
> import ec.Breeder;
> import ec.EvolutionState;
> import ec.Population;
> import ec.util.Parameter;
> import ec.vector.DoubleVectorIndividual;
> /**
> * PSOBreeder.java
> *
>
> <p>The PSOBreeder performs the calculations to determine new  
> particle locations
> and performs the bookkeeping to keep track of personal,  
> neighborhood, and global
> best solutions.
>
> <p><b>Parameters</b><br>
> <table>
> <tr><td valign=top><i>base.</i><tt>debug-info</tt><br>
> <font size=-1>boolean</font></td>
> <td valign=top>(whether the system should display information useful  
> for debugging purposes)<br>
> </td></tr>
>
> </table>
>
> * @author Joey Harrison, Ankur Desai
> * @version 1.0
> */
> public class PSOBreeder extends Breeder
>    {
>    public void setup(EvolutionState state, Parameter base)
>        {
>        // intentionally empty
>        }
>
>    public Population breedPopulation(EvolutionState state)
>        {
> 	for(int s = 0; s < state.population.subpops.length; s++)
> 	    {
> 	    PSOSubpopulation subpop = (PSOSubpopulation)  
> state.population.subpops[s];
> 		
> 	    // update bests
> 	    assignPersonalBests(subpop);
> 	    assignNeighborhoodBests(subpop);
> 	    assignGlobalBest(subpop);
>
> 	    // make a temporary copy of locations so we can modify the  
> current location on the fly
> 	    DoubleVectorIndividual[] tempClone = new  
> DoubleVectorIndividual[subpop.individuals.length];
> 	    System.arraycopy(subpop.individuals, 0, tempClone, 0,  
> subpop.individuals.length);
> 		
> 	    // update particles
> 	    for (int i = 0; i < subpop.individuals.length; i++)
> 		{
> 		DoubleVectorIndividual ind =  
> (DoubleVectorIndividual)subpop.individuals[i];
> 		DoubleVectorIndividual prevInd =  
> (DoubleVectorIndividual)subpop.previousIndividuals[i];
> 		// the individual's personal best
> 		DoubleVectorIndividual pBest =  
> (DoubleVectorIndividual)subpop.personalBests[i];
> 		// the individual's neighborhood best
> 		DoubleVectorIndividual nBest =  
> (DoubleVectorIndividual)subpop.neighborhoodBests[i];
> 		// the individuals's global best
> 		DoubleVectorIndividual gBest =  
> (DoubleVectorIndividual)subpop.globalBest;
> 			
> 		// calculate update for each dimension in the genome
> 		for (int j = 0; j < ind.genomeLength(); j++)
> 		    {
> 		    double velocity = ind.genome[j] - prevInd.genome[j];
> 		    double pDelta = pBest.genome[j] -  
> ind.genome[j];                        // difference to personal best
> 		    double nDelta = nBest.genome[j] -  
> ind.genome[j];                        // difference to neighborhood  
> best
> 		    double gDelta = gBest.genome[j] -  
> ind.genome[j];                        // difference to global best
> 		    double pWeight =  
> state.random[0].nextDouble();                          // weight for  
> personal best
> 		    double nWeight =  
> state.random[0].nextDouble();                          // weight for  
> neighborhood best
> 		    double gWeight =  
> state.random[0].nextDouble();                          // weight for  
> global best
> 		    double newDelta = (velocity + pWeight*pDelta + nWeight*nDelta  
> + gWeight*gDelta) / (1+pWeight+nWeight+gWeight);
> 			
> 		    // update this individual's genome for this dimension
> 		    ind.genome[j] += newDelta * subpop.velocityMultiplier;     //  
> it's obvious if you think about it
> 		    }
> 		
> 		if (subpop.clampRange)
> 		    ind.clamp();
> 		}
> 		
> 	    // update previous locations
> 	    subpop.previousIndividuals = tempClone;
> 	    }
> 	return state.population;
>        }
>
>    public void assignPersonalBests(PSOSubpopulation subpop)
>        {
>        for (int i = 0; i < subpop.personalBests.length; i++)
>            if ((subpop.personalBests[i] == null) ||  
> subpop 
> .individuals[i].fitness.betterThan(subpop.personalBests[i].fitness))
>                subpop.personalBests[i] =  
> (DoubleVectorIndividual)subpop.individuals[i].clone();
>        }
>
>    public void assignNeighborhoodBests(PSOSubpopulation subpop)
>        {
>        for (int j = 0; j < subpop.individuals.length; j++)
>            {
>            DoubleVectorIndividual hoodBest =  
> subpop.neighborhoodBests[j];
>            int start = (j - subpop.neighborhoodSize / 2);
>            if (start < 0)
>                start += subpop.individuals.length;
>
>            for (int i = 0; i < subpop.neighborhoodSize; i++)
>                {
>                DoubleVectorIndividual ind =  
> (DoubleVectorIndividual)subpop.individuals[(start + i) %  
> subpop.individuals.length];
>                if((hoodBest == null) ||  
> ind.fitness.betterThan(hoodBest.fitness))
>                    hoodBest = ind;
>                }
>
>            if (hoodBest != subpop.neighborhoodBests[j])
>                subpop.neighborhoodBests[j] =  
> (DoubleVectorIndividual)hoodBest.clone();
>            }
>        }
>
>    public void assignGlobalBest(PSOSubpopulation subpop)
>        {
>        DoubleVectorIndividual globalBest = subpop.globalBest;
>        for (int i = 0; i < subpop.individuals.length; i++)
>            {
>            DoubleVectorIndividual ind =  
> (DoubleVectorIndividual)subpop.individuals[i];
>            if ((globalBest == null) ||  
> ind.fitness.betterThan(globalBest.fitness))
>                globalBest = ind;
>            }
>        if (globalBest != subpop.globalBest)
>            subpop.globalBest =  
> (DoubleVectorIndividual)globalBest.clone();
>        }
>    }

Top of Message | Previous Page | Permalink

Advanced Options


Options

Log In

Log In

Get Password

Get Password


Search Archives

Search Archives


Subscribe or Unsubscribe

Subscribe or Unsubscribe


Archives

June 2019
April 2019
March 2019
January 2019
December 2018
November 2018
October 2018
July 2018
May 2018
January 2018
December 2017
November 2017
October 2017
September 2017
August 2017
July 2017
June 2017
May 2017
April 2017
March 2017
February 2017
January 2017
December 2016
November 2016
October 2016
September 2016
August 2016
July 2016
June 2016
May 2016
April 2016
March 2016
February 2016
January 2016
December 2015
November 2015
October 2015
September 2015
August 2015
July 2015
June 2015
May 2015
April 2015
March 2015
February 2015
January 2015
December 2014
November 2014
September 2014
August 2014
July 2014
June 2014
May 2014
April 2014
March 2014
February 2014
January 2014
December 2013
November 2013
October 2013
September 2013
August 2013
July 2013
June 2013
May 2013
April 2013
March 2013
February 2013
January 2013
December 2012
November 2012
October 2012
September 2012
July 2012
June 2012
May 2012
April 2012
March 2012
February 2012
January 2012
December 2011
November 2011
October 2011
September 2011
August 2011
July 2011
June 2011
May 2011
April 2011
March 2011
February 2011
January 2011
December 2010
November 2010
October 2010
September 2010
August 2010
July 2010
June 2010
May 2010
April 2010
March 2010
February 2010
January 2010
December 2009
November 2009
October 2009
September 2009
August 2009
July 2009
June 2009
May 2009
April 2009
March 2009
February 2009
January 2009
December 2008
November 2008
October 2008
September 2008
August 2008
July 2008
June 2008
May 2008
April 2008
March 2008
February 2008
January 2008
December 2007
November 2007
October 2007
September 2007
August 2007
July 2007
June 2007
May 2007
April 2007
March 2007
February 2007
January 2007
December 2006
November 2006
October 2006
September 2006
August 2006
July 2006
June 2006
May 2006
April 2006
March 2006
February 2006
January 2006
December 2005
November 2005
October 2005
September 2005
August 2005
July 2005
June 2005
May 2005
April 2005
March 2005
February 2005
January 2005
December 2004
November 2004
September 2004
August 2004
July 2004
June 2004
May 2004
April 2004
March 2004

ATOM RSS1 RSS2



LISTSERV.GMU.EDU

CataList Email List Search Powered by the LISTSERV Email List Manager