December 18, 2002
TiVo Utility Graphs
I just got back from a meeting with Ketan where we talked about where to go from here, now that we have some good results. The basic idea is to compare human awareness (X-axis) to automatic error (Y-axis) as a measure of utility (Z-axis). The intersection of the human and automatic curves should create a line that indicates the point where a human does better than a heuristic policy (and vice versa). The idea now is to make three 2-D graphs of human versus automatic for five content utility distribution (CUD) functions, five time decay values, and five consumption rates. And also a 3-D graph providing an example of the X-Y-Z graph I described above.
Ketan's already made quite a bit of progress on the write up. My job now is to create a program that given the two data sets computes the intersection to create the 2-D graphs. The five existing CUDs are good, but we want to get more data points under 10% human awareness, since that appears to be the point where the human policy always does better. So maybe 1%-15% will be our awareness range. And push up error to as much as +/-50% since +/-25% doesn't always seem to show much difference. I also need to find a good fixed point of CUD and decay for the 3-D graph, which will also be the fixed point used for the five consumption rates.
Posted by josuah at December 18, 2002 7:56 PM UTC+00:00
Trackback Pings
TrackBack URL for this entry:
http://www.wesman.net/cgi-bin/mt/mt-tb.cgi/575
Nami & Kiba - Away From Home
Goodbye Ellie
Evangelion Mini Cooper S Clubman
Married to Christina
Comments
Post a comment
Thanks for signing in, . Now you can comment. (sign out)
(If you haven't left a comment here before, you may need to be approved by the site owner before your comment will appear. Until then, it won't appear on the entry. Thanks for waiting.)