Stop condition for subgradient minimization in dual relaxed (max,+) problem

Authors:
Michail Schlesinger;Evgeniy Vodolazskiy;Nikolai Lopatka
Affiliations:
International Research & Training Center of Information Technologies and Systems of National Academy of Sciences of Ukraine;International Research & Training Center of Information Technologies and Systems of National Academy of Sciences of Ukraine;International Research & Training Center of Information Technologies and Systems of National Academy of Sciences of Ukraine
Venue:
EMMCVPR'11 Proceedings of the 8th international conference on Energy minimization methods in computer vision and pattern recognition
Year:
2011

Citing 3
Cited 0

Convergent Tree-Reweighted Message Passing for Energy Minimization

IEEE Transactions on Pattern Analysis and Machine Intelligence
Diffusion algorithms and structural recognition optimization problems

Cybernetics and Systems Analysis
MAP estimation via agreement on trees: message-passing and linear programming

IEEE Transactions on Information Theory

Quantified Score

Hi-index	0.00

Visualization

Abstract

Subgradient descent methods for minimization of dual linear relaxed labeling problem are analysed. They are guaranteed to converge to the quality of the optimal relaxed labeling, but do not obtain an optimal relaxed labeling itself. Moreover, no stop condition is known for these methods upto now. The stop condition is defined and experimentally compared with the commonly-used stop conditions. The stop condition is defined in a way that when fulfilled a relaxed labeling is simultaneously obtained with arbitrary non-zero difference from the optimal labeling.