ILP-Solver

Description:

Integral linear problems occur in machine learning in the context of structured prediction problems. The resulting problems are very large and need to be solved quickly. We develop fast, flexible, and theoretically grounded algorithms for their solution, with a particular focus on parallelizable GPU algorithms.

Literatur:

• "FastDOG: Fast Discrete Optimization on GPU", Abbas, Ahmed and Swoboda, Paul, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2022 
• "Structured Prediction Problem Archive", Swoboda, Paul and  Andres, Bjoern, Bernard, Florian, Irmai, Jannik, Hornakova, Andrea and Roetzer, Paul and Savchynskyy, Bogdan and Stein, David and Abbas, Ahmed, Arxiv 2022 
• "Efficient Message Passing for 0-1 {ILPs} with Binary Decision Diagrams", Lange, Jan-Hendrik and Swoboda, Paul, Proceedings of the International Conference on Machine Learning (ICML) 2021
• "MAP inference via Block-Coordinate Frank-Wolfe Algorithm", Swoboda, Paul and Kolmogorov, Vladimir, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2019 
• "A Dual Ascent Framework for Lagrangean Decomposition of Combinatorial Problems", Swoboda, Paul and Kuske, Jan and Savchynskyy, Bogdan, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 2017