ICML 2019oral48 citations

Rao-Blackwellized Stochastic Gradients for Discrete Distributions

Runjing Liu, Jeffrey Regier, Nilesh Tripuraneni, Michael Jordan, Jon Mcauliffe

Abstract

We wish to compute the gradient of an expectation over a finite or countably infinite sample space having K $\leq$ $\infty$ categories. When K is indeed infinite, or finite but very large, the relevant summation is intractable. Accordingly, various stochastic gradient estimators have been proposed. In this paper, we describe a technique that can be applied to reduce the variance of any such estimator, without changing its bias{—}in particular, unbiasedness is retained. We show that our technique is an instance of Rao-Blackwellization, and we demonstrate the improvement it yields on a semi-supervised classification problem and a pixel attention task.

BibTeX
@InProceedings{pmlr-v97-liu19c,
  title = 	 {Rao-Blackwellized Stochastic Gradients for Discrete Distributions},
  author =       {Liu, Runjing and Regier, Jeffrey and Tripuraneni, Nilesh and Jordan, Michael and Mcauliffe, Jon},
  booktitle = 	 {Proceedings of the 36th International Conference on Machine Learning},
  pages = 	 {4023--4031},
  year = 	 {2019},
  editor = 	 {Chaudhuri, Kamalika and Salakhutdinov, Ruslan},
  volume = 	 {97},
  series = 	 {Proceedings of Machine Learning Research},
  month = 	 {09--15 Jun},
  publisher =    {PMLR},
  pdf = 	 {http://proceedings.mlr.press/v97/liu19c/liu19c.pdf},
  url = 	 {https://proceedings.mlr.press/v97/liu19c.html},
  abstract = 	 {We wish to compute the gradient of an expectation over a finite or countably infinite sample space having K $\leq$ $\infty$ categories. When K is indeed infinite, or finite but very large, the relevant summation is intractable. Accordingly, various stochastic gradient estimators have been proposed. In this paper, we describe a technique that can be applied to reduce the variance of any such estimator, without changing its bias{—}in particular, unbiasedness is retained. We show that our technique is an instance of Rao-Blackwellization, and we demonstrate the improvement it yields on a semi-supervised classification problem and a pixel attention task.}
}
Rao-Blackwellized Stochastic Gradients for Discrete Distributions · ICML 2019