[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Neural Networks in octave
From: |
Søren Hauberg |
Subject: |
Re: Neural Networks in octave |
Date: |
Sat, 10 Mar 2007 11:25:10 +0100 |
User-agent: |
Thunderbird 1.5.0.10 (X11/20070307) |
Hi,
Carlos Medrano skrev:
http://hauberg.org/wiki/doku.php?id=octave-fann/
I'm the author of this so you can't trust anything I say about this
package ;-)
The FANN bindings work quite well, but FANN has some problems that also
affect the bindings:
*) There is currently no way to get access to the individual weights
in the network. Sometimes you care about this, sometimes you don't.
*) FANN only supports Feedforward Networks which is good enough for
some uses, but sometimes you just need more powerful algorithms.
Other than that Octave-FANN is quite nice. It's fast, it's easy, and you
can do the training in Octave and then later run the network from C
which can be quite handy.
but I would like to know the opinion of some experienced users before choosing
octave or matlab (and his neural network toolbox).
I believe there's also some NN code in octave-forge but I don't know how
well it works. I guess you could also try out "netlab" (just google it)
which is
matlab toolbox for pattern recognition written by Bishop. I haven't
tried it but I assume that it works in octave.
As to if you should use octave or matlab my advise is to use octave
(you're asking at an octave list so this is the expected answer). The
reason is that in the longer run you will need the power that using free
software provides. Matlab might be more handy in the short run, but you
will hit the "propriatory software" wall sooner or later.
Søren