[Paradiseo-help] Help with PSO
Kevin O Mahony
komahony at gmail.com
Mer 3 Juin 21:36:15 CEST 2009
First I'd like to thank you/your group for building such a comprehensive EA
and making it freely available to all. I am currently using
paradiseo-ix86-1.1 in my Masters thesis.
My project compares GA with PSO optimization for use in generating trading
for the intrady foreign exchange market and i've hit a major problem.
can help me.
The GA version behaves as expected, converging nicely, but the PSO version
not converge at all - i'm using the eoStat feature to print out best, mean,
per generation and it does not behave as expected.
The code is identical for both - same fitness function is used for both GA
except for types etc. with the same input data set.
I've used your example GA and binary PSO code for the setup/running of the
I also used a real PSO version again based on example code with the same
I'm working on Debian 4.1.1r21, kernel 2.6.18 with g++ 4.1.2.
Are there any known issues with my environment?
I notice a contrib directory with a new PSO version in there. I was unable
to build it - complaints
of missing definitions for virtual methods. Is this significantly different
to the version in paradiseo-eo
or just a packaging up of existing features?
I've used the same fitness function with PSO code from 1998 by Yuhui Shi, a
Kennedy/Eberhart and i get convergence - but it seems to get stuck very
early on in
maybe a local maximum. The algorithm is also very basic.
I'm under severe time pressure to get a PSO version working so that i can do
comparison with GA - doing both in Paradiseo is required and there are not
PSO frameworks like yours out there to choose from.
I've left out the nitty gritty detail above - if nothing obvious occurs
straight away, i can supply
whatever other details you might need to point me in the direction of a
-------------- section suivante --------------
Une pièce jointe HTML a été nettoyée...
Plus d'informations sur la liste de diffusion Paradiseo-help