Vorträge und Posterpräsentationen (mit Tagungsband-Eintrag):
J. Jancsary, G. Matz, H. Trost:
"An Incremental Subgradient Algorithm for Approximate MAP Estimation in Graphical Models";
Vortrag: NIPS International Workshop on Optimization for Machine Learning,
Whistler (Canada);
10.12.2010; in: "Proc. 3rd International Workshop on Optimization for Machine Learning",
(2010),
6 S.
Kurzfassung englisch:
We present an incremental subgradient algorithm for approximate computation of maximum-a-posteriori (MAP) states in cyclic graphical models. Its most strik- ing property is its immense simplicity: each iteration requires only the solution of a sequence of trivial optimization problems. The algorithm can be equally un- derstood as a degenerated dual decomposition scheme or as minimization of a degenerated tree-reweighted upper bound and assumes a form that is reminiscent of message-passing. Despite (or due to) its conceptual simplicity, it is equipped with important theoretical guarantees and exposes strong empirical performance.
Elektronische Version der Publikation:
http://publik.tuwien.ac.at/files/PubDat_193969.pdf