LP_MP-MRF

Dual decomposition based Markov-Random-Fields solvers.

Available solvers

• SRMP [1]
• MPLP [2]

Also, tightening with cycle inequalities [3] is supported.

Input formats

Input is supported in the UAI format as well as in the hdf5-based opengm format.

Installation

Type git clone https://github.com/pawelswoboda/LP_MP-MRF.git, then cd LP_MP-MRF and initialize dependencies via git submodule update --init --recursive. Then type cmake . to configure and make to build.

References

• [1]: V. Kolmogorov. A New Look at Reweighted Message Passing. IEEE Trans. Pattern Anal. Mach. Intell., 37(5):919–930, 2015.
• [2]: A. Globerson and T. S. Jaakkola. Fixing max-product: Convergent message passing algorithms for MAP LP-relaxations. In NIPS, pages 553–560, 2007.
• [3]: D. Sontag, D. K. Choe, and Y. Li. Efficiently searching for frustrated cycles in MAP inference. In UAI, pages 795–804. AUAI Press, 2012.