Parrallel Network Training with FANN

Steffen Nissen

2009-05-07 20:40:56 UTC

Hi Kerva,

Please move this discussion to the FANN forum
http://leenissen.dk/fann/forum/, as the mailing list have more or less
been decommissioned.

Artificial neural network training algorithms are as such not ideal for
distribution, since they are an iterative process where the result of one
iteration is used in the next iterations.

If the dataset is VERY large, it should, however, be possible to make the
training somewhat parallel. And the candidate neurons for the cascade
algorithm can also be trained separately.

To gain an actual general advantage, I would, however recommend implementing
a new algorithm which trains several networks or something like that.

Best Regards,
Steffen Nissen, MSc
http://www.linkedin.com/in/steffennissen

Post by Jose Luis Kerva
Hello,
i would like to run the training process either on separate 20 linux
stations or on a cluster of those 20 stations. Is it possible to do this
with FANN ? I want to run a _single_ network on this cluster. If source
could has to be modified, would it be much work to do to achieve this?
Thanks in advance.
K
¡Obtén la mejor experiencia en la web! Descarga gratis el nuevo
Internet Explorer 8. http://downloads.yahoo.com/ieak8/?l=mx
------------------------------------------------------------------------------
The NEW KODAK i700 Series Scanners deliver under ANY circumstances! Your
production scanning environment may not be a perfect world - but thanks to
Kodak, there's a perfect scanner to get the job done! With the NEW KODAK
i700
Series Scanner you'll get full speed at 300 dpi even with all image
processing features enabled. http://p.sf.net/sfu/kodak-com
_______________________________________________
Fann-general mailing list
https://lists.sourceforge.net/lists/listinfo/fann-general