Name
Abstract | ||
|---|---|---|
We present a novel dual decomposition approach to MAP inference with highly connected discrete graphical models. Decompositions into cyclic k-fan structured subproblems are shown to significantly tighten the Lagrangian relaxation relative to the standard local polytope relaxation, while enabling efficient integer programming for solving the subproblems. Additionally, we introduce modified update rules for maximizing the dual function that avoid oscillations and converge faster to an optimum of the relaxed problem, and never get stuck in nonoptimal fixed points. |
Year | DOI | Venue |
|---|---|---|
2010 | 10.1007/978-3-642-15558-1_53 | ECCV (3) |
Keywords | Field | DocType |
nonoptimal fixed point,efficient integer programming,novel dual decomposition approach,mrf inference,discrete graphical model,standard local polytope relaxation,k-fan decomposition,cyclic k-fan,dual function,update rule,map inference,lagrangian relaxation,tight lagrangian relaxation,graphical model,oscillations,fixed point | Applied mathematics,Computer science,Markov random field,Polytope,Integer programming,Artificial intelligence,Fixed point,Lagrangian relaxation,Computer vision,Mathematical optimization,Inference,Graphical model,Linear programming relaxation | Conference |
Volume | ISSN | ISBN |
6313 | 0302-9743 | 3-642-15557-X |
Citations | PageRank | References |
8 | 0.45 | 16 |
Authors | ||
3 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Jörg Hendrik Kappes | 1 | 98 | 5.58 |
| Stefan Schmidt | 2 | 154 | 10.01 |
| Christoph Schnörr | 3 | 3025 | 219.34 |