Train and Test Tightness of LP Relaxations in Structured Prediction
Proceedings of the 33rd International Conference on Machine Learning
MetadataShow full item record
Meshi, O., Mahdavi, M., Weller, A., & Sontag, D. (2016). Train and Test Tightness of LP Relaxations in Structured Prediction. Proceedings of the 33rd International Conference on Machine Learning, 48 1776-1785. https://doi.org/10.17863/CAM.242
This is the author accepted manuscript. The final version is available from Microtome Publishing via http://www.jmlr.org/proceedings/papers/v48/meshi16.html
Structured prediction is used in areas such as computer vision and natural language processing to predict structured outputs such as segmentations or parse trees. In these settings, prediction is performed by MAP inference or, equivalently, by solving an integer linear program. Because of the complex scoring functions required to obtain accurate predictions, both learning and inference typically require the use of approximate solvers. We propose a theoretical explanation to the striking observation that approximations based on linear programming (LP) relaxations are often tight on real-world instances. In particular, we show that learning with LP relaxed inference encourages integrality of training instances, and that tightness generalizes from train to test data.
This record's DOI: https://doi.org/10.17863/CAM.242
This record's URL: https://www.repository.cam.ac.uk/handle/1810/256301