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:


Post a comment

July 2013
Sun Mon Tue Wed Thu Fri Sat
  1 2 3 4 5 6
7 8 9 10 11 12 13
14 15 16 17 18 19 20
21 22 23 24 25 26 27
28 29 30 31