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

Emanuele Olivetti olivetti at itc.it
Thu Dec 7 02:57:00 CST 2006


Marek Wojciechowski wrote:
> Comapring to conx.py ffnet is much, much faster and much, much simpler.  
> Faster
> thanks to scipy optimization routines and simpler because simplicity was
> the basic assumption when I created ffnet.
> 
> To prove it:
> I trained a XOR network (2-2-1) in conx with:
...
> This performed 5000 iterations in 34.6 s (and the tolerance has not been  
> reached)
> 
> With ffnet it can be done with:
...
> The above trains nework in 17.9 ms, and the fitness is perfect.
> 
> ffnet is almost 2000 times faster than conx in this example! An the code  
> is simpler.
> 
> However there are some nice testing methods in conx, which can be used in  
> ffnet.

Hi Marek,
Thank you _very_ much for your test. When I need to use ANN is on huge
datasets so performance is a really big issue to me. I'll definitely try
your ffnet very soon on my data. And I hope to see recurrent neural
networks implemented in future releases ;)

Emanuele


More information about the SciPy-user mailing list