repetitive training

[Greg Louis]

Hi, Gary:

Pi pointed out your writeup on his work and I took a look.

I think there is still some doubt possible regarding the benefits of
repetitive training, and it mightn't be a bad idea to mention
    http://www.bgl.nu/bogofilter/training.html
    http://www.bgl.nu/bogofilter/training2.html and
    http://www.bgl.nu/bogofilter/reptrain.html
as giving another perspective on the topic.  Only the first two relate
to training on error, repeatedly, from scratch; the second and third 
also explore what happens if one trains fully on a substantial corpus
and then switches to training on error.

Unlike Pi's results, mine seem to show that there are diminishing
returns, at best, from repetitions of training on error with the same
message sets, though perhaps one to four repetitions may boost accuracy
somewhat.  The question is certainly not yet closed; Pi and I have
discussed methodology and theory, but IMHO what's really needed is more
experimentation.

-- 
| G r e g  L o u i s         | gpg public key: 0x400B1AA86D9E3E64 |
|  http://www.bgl.nu/~glouis |   (on my website or any keyserver) |
|  http://wecanstopspam.org in signatures helps fight junk email. |