NeurIPS 2016poster9 citations

Near-Optimal Smoothing of Structured Conditional Probability Matrices

Moein Falahatgar, Mesrob I Ohannessian, Alon Orlitsky

Abstract

Utilizing the structure of a probabilistic model can significantly increase its learning speed. Motivated by several recent applications, in particular bigram models in language processing, we consider learning low-rank conditional probability matrices under expected KL-risk. This choice makes smoothing, that is the careful handling of low-probability elements, paramount. We derive an iterative algorithm that extends classical non-negative matrix factorization to naturally incorporate additive smoothing and prove that it converges to the stationary points of a penalized empirical risk. We then derive sample-complexity bounds for the global minimizer of the penalized risk and show that it is within a small factor of the optimal sample complexity. This framework generalizes to more sophisticated smoothing techniques, including absolute-discounting.

BibTeX
@inproceedings{NIPS2016_8bdb5058,
 author = {Falahatgar, Moein and Ohannessian, Mesrob I and Orlitsky, Alon},
 booktitle = {Advances in Neural Information Processing Systems},
 editor = {D. Lee and M. Sugiyama and U. Luxburg and I. Guyon and R. Garnett},
 pages = {},
 publisher = {Curran Associates, Inc.},
 title = {Near-Optimal Smoothing of Structured Conditional Probability Matrices},
 url = {https://proceedings.neurips.cc/paper_files/paper/2016/file/8bdb5058376143fa358981954e7626b8-Paper.pdf},
 volume = {29},
 year = {2016}
}
Near-Optimal Smoothing of Structured Conditional Probability Matrices · NeurIPS 2016