Title :
On the convergence of the iterative “pseudo likelihood” maximization algorithm
Author_Institution :
IRISA-INRIA, Rennes
fDate :
March 31 2008-April 4 2008
Abstract :
We consider a powerful iterative inference algorithm which has recently appeared in the literature. In this paper, we refer to this algorithm as iterative "pseudo likelihood" maximization (IPLFM) algorithm. We give a connection between this algorithm and the problem of Bethe free energy minimization and prove several important results concerning its fixed points and its convergence properties.
Keywords :
convergence of numerical methods; iterative methods; minimisation; Bethe free energy minimization; convergence properties; iterative pseudo likelihood maximization algorithm; Bismuth; Convergence of numerical methods; Gradient methods; Inference algorithms; Iterative algorithms; Iterative methods; Minimization methods; Sufficient conditions; MAP estimation; convergence of numerical methods; iterative methods;
Conference_Titel :
Acoustics, Speech and Signal Processing, 2008. ICASSP 2008. IEEE International Conference on
Conference_Location :
Las Vegas, NV
Print_ISBN :
978-1-4244-1483-3
Electronic_ISBN :
1520-6149
DOI :
10.1109/ICASSP.2008.4518456
