pawelswoboda/LP_MP

Solving LPs with convergent message passing

LP_MP

LP_MP is a C++ framework for developing scalable dual (Lagrangean) decomposition based algorithms solvers for a wide range of LP-relaxations to discrete optimization problems. For a theoretical introduction to the techniques used and the class of problems that can be optimized see [1].

Solvers

Solvers are provided in separate projects and include

• Discrete graphical models,
• Multicut,
• Graph matching,
• Discrete tomography.

Optimization techniques

Optimization techniques include

• Messsage passing [1],
• Subgradient ascent with a proximal bundle method based on the Frank-Wolfe algorithm [2], Vladimir Kolmogorov's original implementation.
• An interface to external solvers is provided by DD_ILP.

Installation

Type git clone https://github.com/pawelswoboda/LP_MP.git for downloading, then cd LP_MP and git submodule update --init for downloading dependenciesand finallycmake` for building.

Prerequisites:

• Clang 5.0 or GCC 7.0 upwards for C++17 compatibility.

References

• [1]: P. Swoboda, J. Kuske and B. Savchynskyy. A Dual Ascent Framework for Lagrangean Decomposition of Combinatorial Problems. In CVPR 2017.
• [2]: P. Swoboda and V. Kolmogorov. MAP inference via Block-Coordinate Frank-Wolfe Algorithm. arXiv.