Title :
Generalized linear coordinate-descent message-passing for convex optimization
Author :
Zhang, Guoqiang ; Heusdens, Richard
Author_Institution :
Dept. of Mediamatics, Delft Univ. of Technol., Delft, Netherlands
Abstract :
In this paper we propose a generalized linear coordinate-descent (GLiCD) algorithm for a class of unconstrained convex optimization problems. The considered objective function can be decomposed into edge-functions and node-functions of a graphical model. The messages of the GLiCD algorithm are in a form of linear functions, as compared to the min-sum algorithm of which the form of messages depends on the objective function. Thus, the implementation of the GLiCD algorithm is much simpler than that of the min-sum algorithm. A theorem is stated according to which the algorithm converges to the optimal solution if the objective function satisfies a diagonal-dominant condition. As an application, the GLiCD algorithm is exploited in solving the averaging problem in sensor networks, where the performance is compared to that of the min-sum algorithm.
Keywords :
convex programming; graph theory; message passing; convex optimization; edge-functions; generalized linear coordinate-descent message-passing; graphical model; min-sum algorithm; node-functions; Algorithm design and analysis; Convergence; Convex functions; Graphical models; Linear programming; Optimization; Vectors; Convex optimization; coordinate decent; message passing;
Conference_Titel :
Acoustics, Speech and Signal Processing (ICASSP), 2012 IEEE International Conference on
Conference_Location :
Kyoto
Print_ISBN :
978-1-4673-0045-2
Electronic_ISBN :
1520-6149
DOI :
10.1109/ICASSP.2012.6288302