On Wed, 2003-07-09 at 11:29, Muenz, Michael wrote: > From: "Simon Byrnand" <[EMAIL PROTECTED]>: > > > Hi ML, > > > > > > I've set auto_learn_threshold_spam to 14.0. > > > My Q. is, when I got a Spam e.g. from @valodata.com > > > it gets about 10.0, then auto_learn will not learn from this, > > > so I set the domain to manual blacklist that it gets +100. > > > Will then SA learn from this Spam ? I think so, but I'd > > > like to hear a "definitely yes" :-) > > > THX > > > > No it won't. > > > > Manual blacklist/whitelist entries and Bayes scores don't contribute > > towards triggering the autolearn threshold. > > hmmmm .. my problem is that I use SA on a relayer and > don't want to feed SA manually with sa-learn. But with only > auto_learn enabled I catch plenty of Spam and not much Ham. > Should I set the auto_learn value for nonspam to "0" rather than "-2" ? > THX > > - Michael > > I'm not quite on the same setup but manually learning ham from my own mailboxes is a pain in the ass so I don't do it.
I looked through 3 months worth of spam and couldn't find one below 1 so I picked a value below that. I run mine at 0.6 because at that it autolearns as much ham as it does spam, I selected the autolearn threshold to try and balance out the quantities learned. This isn't a perfect solution because it means bayes never gets to learn my spammiest looking ham, so bayes isn't as efficient as it might be, but it seems to work pretty good and shows no signs of degrading so far. -- Yorkshire Dave -- Scanned by MailScanner at wot.no-ip.com ------------------------------------------------------- This SF.Net email sponsored by: Parasoft Error proof Web apps, automate testing & more. Download & eval WebKing and get a free book. www.parasoft.com/bulletproofapps _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
