views:

131

answers:

4

I'm a College student (Economics) and I want to program some monetary models using Neural Networks. I want those models to be able to predict future values of some variables using economic data, but I really don't know how to "model" the program itself. Is there any good Python module for that? I mean, a module for NN and a module for economic analysis?

P.S.: I'm using Python 3.x, but I can switch to 2.7.x if needed...

+2  A: 

Google yields at least four different Python neural network implementations; in particular, bpnn.py looks good just for its simplicity.

Or were you looking for an explanation of neural networks?

katrielalex
That site is great! Just was I was looking for... Thank you guys! But there is a little (big) problem: the posts dates from 2006. Two of the links are not working ( the financial example: that was great, I suppose!) The other two: are they compatible with Python 2.7.x? (no way with 3.x!!!). I don't want to "translate" the entire libraries. Anyway, the link is very useful. Thanks!
Hannibal Wagner
Most 2.x libraries are forward compatible with later versions. Python 2.6 is the most popular version by far; 3.x will take a while before the bulk of libraries support it (though some small simple old ones may never).
Nick T
+2  A: 

I've played with ffnet a little. PS - It was a pain to install.

"Feed-forward neural network for python"
http://pypi.python.org/pypi/ffnet/0.6

Louis
Nice! But only for Feed-forward networks. Thank you!
Hannibal Wagner
+1  A: 

There is also PyBrain. I didn't work with any library yet, but had some time to investigate the documentation. It looks like PyBrains has the simplest interface, compared to the available feature set.

bgbg
A: 

For large neural networks, you might want to consider GPU-accelerated libraries. Our own library CUV comes to mind, as well as for example theano (try google for that "theano library", I cannot yet post more than one link). CUV has python bindings, theano actually generates C++/CUDA code.

hannes