[Spambayes-checkins] website background.ht,1.16,1.17
Anthony Baxter
anthonybaxter at users.sourceforge.net
Thu Sep 11 01:04:25 EDT 2003
Update of /cvsroot/spambayes/website
In directory sc8-pr-cvs1:/tmp/cvs-serv12193
Modified Files:
background.ht
Log Message:
ascribe blame where blame is due. <wink>
Thanks to Gary (and Tim) for updating my view of history.
Index: background.ht
===================================================================
RCS file: /cvsroot/spambayes/website/background.ht,v
retrieving revision 1.16
retrieving revision 1.17
diff -C2 -d -r1.16 -r1.17
*** background.ht 4 Aug 2003 02:49:59 -0000 1.16
--- background.ht 11 Sep 2003 07:04:22 -0000 1.17
***************
*** 136,144 ****
with an 'unsure' answer.
</P>
! <P>
! These two central limit
! approaches were dropped after Tim, Gary and Rob Hooft produced a combining
! scheme using <a href="http://mathworld.wolfram.com/Chi-SquaredDistribution.html">chi-squared probabilities</a>. This is now the default combining
! scheme. </p>
<P>Chi-combining is similar to the central limit approaches, but it doesn't
have the annoying training problems that central limit approaches suffered
--- 136,149 ----
with an 'unsure' answer.
</P>
! <P>The central limit approaches were dropped after Gary suggested a
! novel but well-founded combining scheme using
! <a href="http://mathworld.wolfram.com/Chi-SquaredDistribution.html">chi-squared
! probabilities</a>.
! Tim implemented it, including a unique routine to compute
! -2*sum(ln p_i) without logarithms and without underflow worries.
! Many, many testers reported excellent results. This is now the
! default combining scheme. Some end cases were improved when
! Rob Hooft discovered a cleaner way to combine the internal overall ham
! and spam scores, which improved detection of the "middle ground".</P>
<P>Chi-combining is similar to the central limit approaches, but it doesn't
have the annoying training problems that central limit approaches suffered
More information about the Spambayes-checkins
mailing list