Learning Maximum-A-Posteriori Perturbation Models for Structured Prediction in Polynomial Time

Asish Ghoshal, Jean Honorio
Proceedings of the 35th International Conference on Machine Learning, PMLR 80:1754-1762, 2018.

Abstract

MAP perturbation models have emerged as a powerful framework for inference in structured prediction. Such models provide a way to efficiently sample from the Gibbs distribution and facilitate predictions that are robust to random noise. In this paper, we propose a provably polynomial time randomized algorithm for learning the parameters of perturbed MAP predictors. Our approach is based on minimizing a novel Rademacher-based generalization bound on the expected loss of a perturbed MAP predictor, which can be computed in polynomial time. We obtain conditions under which our randomized learning algorithm can guarantee generalization to unseen examples.

Cite this Paper


BibTeX
@InProceedings{pmlr-v80-ghoshal18a, title = {Learning Maximum-A-Posteriori Perturbation Models for Structured Prediction in Polynomial Time}, author = {Ghoshal, Asish and Honorio, Jean}, booktitle = {Proceedings of the 35th International Conference on Machine Learning}, pages = {1754--1762}, year = {2018}, editor = {Dy, Jennifer and Krause, Andreas}, volume = {80}, series = {Proceedings of Machine Learning Research}, month = {10--15 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v80/ghoshal18a/ghoshal18a.pdf}, url = {https://proceedings.mlr.press/v80/ghoshal18a.html}, abstract = {MAP perturbation models have emerged as a powerful framework for inference in structured prediction. Such models provide a way to efficiently sample from the Gibbs distribution and facilitate predictions that are robust to random noise. In this paper, we propose a provably polynomial time randomized algorithm for learning the parameters of perturbed MAP predictors. Our approach is based on minimizing a novel Rademacher-based generalization bound on the expected loss of a perturbed MAP predictor, which can be computed in polynomial time. We obtain conditions under which our randomized learning algorithm can guarantee generalization to unseen examples.} }
Endnote
%0 Conference Paper %T Learning Maximum-A-Posteriori Perturbation Models for Structured Prediction in Polynomial Time %A Asish Ghoshal %A Jean Honorio %B Proceedings of the 35th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2018 %E Jennifer Dy %E Andreas Krause %F pmlr-v80-ghoshal18a %I PMLR %P 1754--1762 %U https://proceedings.mlr.press/v80/ghoshal18a.html %V 80 %X MAP perturbation models have emerged as a powerful framework for inference in structured prediction. Such models provide a way to efficiently sample from the Gibbs distribution and facilitate predictions that are robust to random noise. In this paper, we propose a provably polynomial time randomized algorithm for learning the parameters of perturbed MAP predictors. Our approach is based on minimizing a novel Rademacher-based generalization bound on the expected loss of a perturbed MAP predictor, which can be computed in polynomial time. We obtain conditions under which our randomized learning algorithm can guarantee generalization to unseen examples.
APA
Ghoshal, A. & Honorio, J.. (2018). Learning Maximum-A-Posteriori Perturbation Models for Structured Prediction in Polynomial Time. Proceedings of the 35th International Conference on Machine Learning, in Proceedings of Machine Learning Research 80:1754-1762 Available from https://proceedings.mlr.press/v80/ghoshal18a.html.

Related Material