How to Compute Primal Solution from Dual One in MAP Inference in MRF?
The result's identifiers
Result code in IS VaVaI
<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68407700%3A21230%2F11%3A00187153" target="_blank" >RIV/68407700:21230/11:00187153 - isvavai.cz</a>
Result on the web
—
DOI - Digital Object Identifier
—
Alternative languages
Result language
angličtina
Original language name
How to Compute Primal Solution from Dual One in MAP Inference in MRF?
Original language description
In LP relaxation of MAP inference in Markov random fields (MRF), the primal LP maximizes the MAP objective over relaxed labelings (pseudomarginals) and the dual LP minimizes an upper bound on the true MAP solution by reparameterizations. Having solved the dual~LP, we have no direct access to the corresponding primal solution. We propose a simple way to compute an optimal primal solution from an optimal dual solution. Precisely, we given an algorithm that either shows that the upper bound for a given problem can be further decreased by reparameterizations (i.e., it is not dual-optimal) or computes the corresponding optimal relaxed labeling. This is done by first removing inactive dual constraints and then solving the resulting feasibility problem by a very simple message-passing algorithm, sum-product diffusion.
Czech name
—
Czech description
—
Classification
Type
J<sub>x</sub> - Unclassified - Peer-reviewed scientific article (Jimp, Jsc and Jost)
CEP classification
JD - Use of computers, robotics and its application
OECD FORD branch
—
Result continuities
Project
—
Continuities
Z - Vyzkumny zamer (s odkazem do CEZ)
Others
Publication year
2011
Confidentiality
S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů
Data specific for result type
Name of the periodical
Control Systems and Computers
ISSN
0130-5395
e-ISSN
—
Volume of the periodical
2011
Issue of the periodical within the volume
2
Country of publishing house
UA - UKRAINE
Number of pages
8
Pages from-to
86-93
UT code for WoS article
—
EID of the result in the Scopus database
—