|
Archives
|
Snowball FightBy conner_bw on 2008-12-07 12:20:16 Just a little update to say that paid work, unrelated to Sux0r, is taking priority until the Christmas holidays. Before I disappear for a few weeks, I've made a few changes to the Naive Bayesian categorization algorithm. What I did was change the "fake probability" condition in an attempt to compensate for skewed training sets. The discussion is here. The code can be obtained via CVS. I'm running the changes on the demo site (i.e. the site you are currently reading) and am getting decent results. Placebo, or math? Code review welcome. Reply Tags: Cvs, Naive Bayesian Categorizer
No comments |
Creative Commons Attribution 3.0. Contributors should be attributed by full name or nickname.