[SciPy-user] feed-forward neural network for python

[Conor Robinson]

Marek,

This looks really good, and seems to be along the lines of stuff I've
been working on.  I like the fact that your using fortran to do some
of the heavy computation, especially the GA.  I'd like to contribute.

I find that the rprop algorithm is the best for real-world-noisy data
(besides GAs), and this could be an easy add. I have a python
implementation that uses scipy to run rprop, there is definite room
for improvement, I'm mainly using scipy.where() and wrote the function
directly from the original rprop paper (i can send you this code if
you'd like it).  However, it may be faster to use fortran here as
well. The original c version or rprop is here
http://www.koders.com/c/fid1B403F9BA6B2BAA1C7CEDD1C680F8AAB2819B44E.aspx

I also think that it would be great to have two versions of this.  One
for general usage and one for altivec and sse 64bit ibm and intel
chips, which at this stage of the project would be easy to modify
http://developer.apple.com/documentation/Performance/Conceptual/Accelerate_sse_migration/index.html
gcc 4.0.3 does some auto optimization with the right flags and I've
found you can increase the speed by factors of 10 or more in many
cases.  What type of platforms do you have access to?  The CHUD tools
on OS X are very effective and might be useful
(http://developer.apple.com/tools/performance/) to find optimizations
relevant to all platforms.

conor


On 12/6/06, Emanuele Olivetti <olivetti at itc.it> wrote:
> Marek Wojciechowski wrote:
> > Hi!
> > I released feed-forward neural network for python project at sourceforge
> > (ffnet). I'm announcing it here because it depends on numpy/scipy tandem,
> > so you, folks, are potential users/testers :)
> >
> > If anyone is interested please visit ffnet.sourceforge.net (and then post
> > comments if any...)
>
> Hi Marek,
> thank you for you ANN implementation. I'm currently interested in recurrent
> neural networks so I hope you'll add them too in near future ;)
>
> Anyway, for those interested in ANN you can check conx.py module of
> PyRobotics project. It's just one python file (182Kb of python source!)
> inside a much bigger project. Conx.py is self contained and unfortunately
> it needs Numeric and not numpy :((
>
> What about mixing ffnet and conx.py? ;)
>
> Some links:
> http://www.pyrorobotics.org/  - main site
> http://www.pyrorobotics.org/?page=PyroModuleNeuralNetworks   -  conx module introduction
> http://cvs.cs.brynmawr.edu/cgi-bin/viewcvs.cgi/pyrobot/brain/conx.py - actual conx.py code
> http://www.pyrorobotics.org/?page=Building_20Neural_20Networks_20using_20Conx - conx example code
>
> HTH
>
> Emanuele
>
> P.S.: both ffnet and conx.py are released under GPL. Good!
> _______________________________________________
> SciPy-user mailing list
> SciPy-user at scipy.org
> http://projects.scipy.org/mailman/listinfo/scipy-user
>