Advances and Applications in Statistics
Volume 15, Issue 2, Pages 101 - 113
(April 2010)
|
|
CONVERGENCE OF NON-HOMOGENEOUS VERSIONS OF THE MCEM AND StEM ALGORITHMS
C. C. Y. Dorea and D. S. B. Martins Neto
|
Abstract: When the expectation of the augmented log-posterior density is not simple to compute, the use of EM algorithms require further stochastic approximations. In this note, non-homogeneous stochastic versions of the EM algorithms are proposed and its performance is compared to that of MCEM and StEM algorithms. Ergodic properties of associated Markov chains are also derived. |
Keywords and phrases: EM algorithm, MCEM, StEM, uniform ergodicity. |
|
Number of Downloads: 341 | Number of Views: 1269 |
|