Name
Abstract | ||
|---|---|---|
Given a directed graphical model with binary-valued hidden nodes and real-valued noisy observations, consider deciding upon the maximum a-posteriori (MAP) or the maximum posterior-marginal (MPM) assign- ment under the restriction that each node broadcasts only to its children exactly one single-bit message. We present a variational formulation, viewing the processing rules local to all nodes as degrees-of-freedom, that minimizes the loss in expected (MAP or MPM) performance subject to such online communication constraints. The approach leads to a novel message-passing algorithm to be executed offline , or before observations are realized, which mitigates the performance loss by iteratively cou- pling all rules in a manner implicitly driven by global stati stics. We also provide (i) illustrative examples, (ii) assumptions that g uarantee conver- gence and efficiency and (iii) connections to active researc h areas. |
Year | Venue | Keywords |
|---|---|---|
2005 | NIPS | graphical model,degree of freedom |
Field | DocType | Citations |
Convergence (routing),Mathematical optimization,Coupling,Computer science,Inference,Theoretical computer science,Artificial intelligence,Graphical model,Machine learning,Global statistics | Conference | 8 |
PageRank | References | Authors |
0.79 | 5 | 2 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| O. Patrick Kreidl | 1 | 66 | 5.17 |
| Alan S. Willsky | 2 | 7466 | 847.01 |