[mythtv] Collaborative filtering and recommendation systems

[f-myth-users at media.mit.edu]

Every so often, someone mentions making some sort of recommendation
system for Myth, and sometimes they build it (e.g., Brad Templeton's
MythWish).

There was a recent discussion about this (April 17-18) started by
Erich Nachbar in the "Paid for Plugin dev?" thread (maybe someone
should link that thread to this one so Gossamer users can find this
from that, but I figured I'd start a new thread for people who weren't
specifically interested in that particular one).

IEEE Spectrum just published a short article about building such
systems at http://www.spectrum.ieee.org/print/8788, which gives a
quick overview of two popular ways people go about it.  Note that
it also mentions that 100 million ratings fit nicely in a typical
laptop's RAM, e.g., that for the scales at which Myth's user base
operates, recommendation systems shouldn't necessarily be hugely
resource-intensive.

The domain is also suitably close to serve as inspiration---the IEEE
article is talking about the Netflix $1M movie-recommendation challenge.
It's unfortunately pretty light on actual technical details, but there's
a large literature---you can go look 'em up. :)