Maximum persistency via iterative relaxed inference with graphical models Journal Article

Author(s): Shekhovtsov, Alexander; Swoboda, Paul; Savchynskyy, Bogdan
Article Title: Maximum persistency via iterative relaxed inference with graphical models
Affiliation IST Austria
Abstract: We consider the NP-hard problem of MAP-inference for undirected discrete graphical models. We propose a polynomial time and practically efficient algorithm for finding a part of its optimal solution. Specifically, our algorithm marks some labels of the considered graphical model either as (i) optimal, meaning that they belong to all optimal solutions of the inference problem; (ii) non-optimal if they provably do not belong to any solution. With access to an exact solver of a linear programming relaxation to the MAP-inference problem, our algorithm marks the maximal possible (in a specified sense) number of labels. We also present a version of the algorithm, which has access to a suboptimal dual solver only and still can ensure the (non-)optimality for the marked labels, although the overall number of the marked labels may decrease. We propose an efficient implementation, which runs in time comparable to a single run of a suboptimal dual solver. Our method is well-scalable and shows state-of-the-art results on computational benchmarks from machine learning and computer vision. IEEE
Keywords: Energy minimization; Discrete optimization; graphical models; LP relaxation; partial optimality; persistency; WCSP
Journal Title: IEEE Transactions on Pattern Analysis and Machine Intelligence
Volume: PP
Issue 99
ISSN: 0162-8828
Publisher: IEEE  
Date Published: 2017-07-24
Start Page: 1
End Page: 1
Sponsor: Austrian Science Fund; German Research Foundation DFG; European Research Council
DOI: 10.1109/TPAMI.2017.2730884
Open access: yes (repository)
IST Austria Authors
Related IST Austria Work