On Sat, July 21, 2007 17:46, C Stephan wrote: > I understand SpamBayes works best when trained on roughly equal > numbers of ham & spam messages. After initial training, I've been > training it only on messages it got wrong or was unsure > about. Should I also be including some (or all) correctly-labeled > ham & spam in the training? Thanks.
You train the same way as most people do. (I think) You don't have to train correctly-labeled ham & spam. Except perhaps when your traing db has a very unequal number of ham vs spam. And even then... -- Amedee _______________________________________________ SpamBayes@python.org http://mail.python.org/mailman/listinfo/spambayes Check the FAQ before asking: http://spambayes.sf.net/faq.html
