Jump to content

Talk:Ordered subset expectation maximization

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

OSEM is an iterative method and not an algorithm. Algorithms terminate with a solution in a finite number of steps. Iterative methods yield sequences some subsequence of which converges in some sense.

Nemirovskii et alia has the best approximation-bound for finite step approximation error (known to me), and it's bad because of the problems of controlling the Lipschitz modulus of uniform continuity for sums. Kiefer.Wolfowitz (talk) 16:23, 21 July 2010 (UTC)[reply]

Assessment comment

[edit]

The comment(s) below were originally left at Talk:Ordered subset expectation maximization/Comments, and are posted here for posterity. Following several discussions in past years, these subpages are now deprecated. The comments may be irrelevant or outdated; if so, please feel free to remove this section.

EM optimizes a function, like the log-likelihood function and this function is responsible for the quality of the estimator. Thus it is not clear why the EM algorithm should improve on Filtered Back projection. Any method which optimizes the same function should do as well, no ? Stephanechretien (talk) 19:30, 21 November 2008 (UTC)[reply]

Last edited at 19:30, 21 November 2008 (UTC). Substituted at 03:17, 3 May 2016 (UTC)