I haven't yet full absorbed the details of this, so I can't really say if this is a "huh." or a "hm!" or a "meh.", but there's a line in the abstract of the attached paper: "We demonstrate that the expected Guesswork for a source conditioned to create words in the typical set grows, with word length, at a lower exponential rate than that of the uniform approximation, suggesting that the use of the approximation is ill-advised."
As I'm reading that, it looks like it might or might not be an interesting conclusion, and I'd be interested in hearing what way some of the smart kids see it.
The Article is from MIT news.
The Paper is linked in the article, but maybe you want to cut straight to the meat of the matter.
0 Replies - 1621 Views - Last Post: 14 August 2013 - 10:08 PM
Page 1 of 1