On Fri, Apr 8, 2011 at 3:24 PM, Eric Charles <[email protected]> wrote:
>
>
>
> Hi Vicki,
> So, will the end user still need to send to spam@.../notspam@... during the
> training session ?

More advanced algorithms may require cross-validation to limit
over-fitting. Typically, it's better to extract the features and then
work with just the numbers. So, I'd be happy to have a Mime4J module
capable of extracting a configurable feature vector by parsing a mail
without worrying too much about wiring this in. I'd find running
against a couple of mailboxes much better than the whole
spam@.../notspam@ stuff.

Robert

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to