NN module(s) for Perl

0 views
Skip to first unread message

gold...@my-dejanews.com

unread,
Apr 27, 1999, 3:00:00 AM4/27/99
to
Is there a NN code (defining a net, training etc) implemented as Perl module?
CPAN couldn't find me one.

peter

-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own

Q Modnar

unread,
Apr 28, 1999, 3:00:00 AM4/28/99
to
Bruce McPherson <bru...@xtra.co.nz> wrote:

> gold...@my-dejanews.com wrote:
>>
>> Is there a NN code (defining a net, training etc) implemented as Perl module?
>> CPAN couldn't find me one.

> I don't think it would be easy to do a NN in Perl as Perl is a very high
> level script language, your best bet would be Java or JavaScript.

You're a pessimist! Perl is more than capable of doing the job...albeit about
20-30% slower than a compiled language (there is a perl compiler though), and
I bet much faster than javascript! Believe it or not but I've even seen
neural networks implemented in zsh! I'd suggest that the original poster get
some C code and do the conversion to a perl module herself/himself....would
make a good learning project.


Bruce McPherson

unread,
Apr 29, 1999, 3:00:00 AM4/29/99
to

gold...@my-dejanews.com

unread,
Apr 29, 1999, 3:00:00 AM4/29/99
to

> >> Is there a NN code (defining a net, training etc) implemented as Perl
module?
> >> CPAN couldn't find me one.
>
> > I don't think it would be easy to do a NN in Perl as Perl is a very high
> > level script language, your best bet would be Java or JavaScript.
>
> You're a pessimist! Perl is more than capable of doing the job...albeit about
> 20-30% slower than a compiled language (there is a perl compiler though), and
> I bet much faster than javascript! Believe it or not but I've even seen
> neural networks implemented in zsh! I'd suggest that the original poster get
> some C code and do the conversion to a perl module herself/himself....would
> make a good learning project.

The original poster (that's me :) is quite lazy and hates the thought of
having to convert C/C++ code to perl module ;). But you are right about that
learning project stuff... hmm... Also, why do you think it will be slower
than 'compiled language'? The module would be a C code already precompiled
and the NN processing should run no slower and no faster than any other NN
simulator written in C/C++. And the perl offers flexibility I need to
preprocess training patterns, can process a lot of different input sources
and to me seems ideal for implementing NNs. It is a bit strange nobody so far
has put together a module for this.

Peter

G.L. Grobe

unread,
Apr 29, 1999, 3:00:00 AM4/29/99
to
It's correct that Perl could easily do the job. I've done this same thing with Tcl.
What I've ended up with is a high-level scripting language extension to Tcl that
enables me to build various networks much quicker and plenty of flexibility. The
module is coded in C++, but I had to be careful on how much processing the scripts
were allowed to do and how many calls were made from the Tcl script. And of course,
after you code your first module, then the second, you always find more ways to
keep on optimizing.

I've got some example Tcl scripts using these new command implentations at
www.neuroquest.com under the Documentation link. I hope this gives you some ideas.
You could even grab this lib and link in a tcl module to perl if you wished.


John Smith

unread,
May 3, 1999, 3:00:00 AM5/3/99
to
There is a Perl module for Neural Nets, not sure how flexible though,
more information can be found in
http://www.tertius.net.au/lists/perl-ai/links.html

G.L. Grobe <gary....@aspentech.com> wrote in message
news:3728C186...@aspentech.com...

Reply all
Reply to author
Forward
0 new messages