DocumentCode :
3467535
Title :
Getting Feasible Variable Estimates from Infeasible Ones: MRF Local Polytope Study
Author :
Savchynskyy, Bogdan ; Schmidt, Signe
Author_Institution :
Univ. of Heidelberg, Heidelberg, Germany
fYear :
2013
fDate :
2-8 Dec. 2013
Firstpage :
267
Lastpage :
274
Abstract :
This paper proposes a method for the construction of approximate feasible primal solutions from infeasible ones for large-scale optimization problems possessing certain separability properties. Whereas the infeasible primal estimates can typically be produced from (sub-) gradients of the dual function, it is often not easy to project them to the primal feasible set, since the projection itself has a complexity comparable to the complexity of the initial problem. We propose an alternative efficient method to obtain feasibility and show that its properties influencing the convergence to the optimum are similar to the properties of the Euclidean projection. We apply our method to the local polytope relaxation of inference problems for Markov Random Fields and discuss its advantages over existing methods.
Keywords :
Markov processes; computational geometry; convergence; gradient methods; optimisation; random processes; relaxation theory; Euclidean projection; MRF local polytope; Markov random fields; approximate feasible primal solutions; convergence; dual function subgradients; feasible variable estimates; infeasible primal estimates; inference problems; large-scale optimization problems; local polytope relaxation; primal feasible set; problem complexity; separability properties; Approximation methods; Conferences; Convergence; Entropy; Linear programming; Optimization; Vectors; LP relaxation; MAP inference; duality gap; feasible primal solution; graphical models; markov random fields;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Computer Vision Workshops (ICCVW), 2013 IEEE International Conference on
Conference_Location :
Sydney, NSW
Type :
conf
DOI :
10.1109/ICCVW.2013.43
Filename :
6755908
Link To Document :
بازگشت