From: Matt Mower <matt.mower@...>
Date: 2005-04-12T16:47:47+09:00
Subject: Re: [ANN| Bayesian Classification for Ruby

On Apr 12, 2005 4:32 AM, "Pe�a, Botp" <botp@delmonte-phil.com> wrote:
> Matt Mower [mailto:matt.mower@gmail.com] wrote:
> 
> #> http://rubyforge.org/projects/classifier/
> #>
> #
> #;-)
> #
> #I ported the Reverend bayesian classifier from Python to Ruby
> #over the weekend.  If only I'd waited ;-)
> 
> did you finished the port? Are you using it?
> 

Yes I finished porting the code and it seems to work for the Robinson
method although what I haven't done yet is verify that it produces the
same results as Reverend, i.e. have I ported it properly?  Using the
Robinson-Fisher method doesn't seem to work so I think I need to go
back and look at my implementation of that algorithm again.

Here is Lucas' example using Bishop:

require 'bishop'

# The block passed here is actually mimicing the default behaviour however
# I add it here to show how the probability combiner algorithm is replaced,
# e.g. with Bishop::robinson_fisher
b = Bishop::Bayes.new { |probs,ignore| Bishop::robinson( probs, ignore ) }

b.train( "interesting", "here are some good words. I hope you love them" )
b.train( "uninteresting", "here are some bad words, I hate you" )
b.guess( "I hate bad words and you" ).each { |c| puts c.join( " = " ) }

outputs

uninteresting = 0.9999

Is there anyone with some python savvy who'd like to collaborate on testing?

M

-- 
Matt Mower :: http://matt.blogs.it/