4PSA Spam Guardian has an bayesian spam filtering that has the ability to learn mails when one feeds it with equal number of SPAM mails as well as valid mails. In bayesian terms the SPAM mails are termed as "junk" mails and valid mails are called "ham" mails.To train the bayesian database to recognise SPAM properly follow the below mentioned steps.
Set "leave message
on server" for the account from the Outlook/outlook express or VPOP3 (if using it) and set
it to 1day minimum.

Log into webmail
for the account that is receiving too many SPAM using the Link http://webmail.yourdomain.com (please replace yourdomain.com with your domain name).

Create two folders namely “junk_learn”
“ham_learn” by clicking on “Folders” and from choose action select
“Create Folder”



Now periodically
ask the users to move the spam mails to “junk_learn”
and copy the valid mails to “ham_learn”. (Please be careful that only
copy the valid mails to the “ham_learn” and do not move, otherwise the
mails will not be downloaded). Please remember to feed the bayesian filter with equal amount of "HAM" as well as "JUNK" mails. After 50 "JUNK" and "HAM" mails each the filter would recognize what is SPAM and what is valid.