A Stochastic Path Integral Differential EstimatoR Expectation Maximization Algorithm

Abstract

The Expectation Maximization (EM) algorithm is of key importance for inference in latent variable models including mixture of regressors and experts, missing observations. This paper introduces a novel EM algorithm, called {\tt SPIDER-EM}, for inference from a training set of size $n$, $n \gg 1$. At the core of our algorithm is an estimator of the full conditional expectation in the {\sf E}-step, adapted from the stochastic path integral differential estimator ({\tt SPIDER}) technique. We derive finite-time complexity bounds for smooth non-convex likelihood: we show that for convergence to an $\epsilon$-approximate stationary point, the complexity scales as $K_{Opt} (n,\epsilon )={\cal O}(\epsilon^{-1})$ and $K_{CE}( n,\epsilon ) = n+ \sqrt{n} {\cal O}( \epsilon^{-1} )$, where $K_{Opt}( n,\epsilon )$ and $K_{CE}(n, \epsilon )$ are respectively the number of {\sf M}-steps and the number of per-sample conditional expectations evaluations. This improves over the state-of-the-art algorithms. Numerical results support our findings.

Cite

Text

Fort et al. "A Stochastic Path Integral Differential EstimatoR  Expectation Maximization Algorithm." Neural Information Processing Systems, 2020.

Markdown

[Fort et al. "A Stochastic Path Integral Differential EstimatoR  Expectation Maximization Algorithm." Neural Information Processing Systems, 2020.](https://mlanthology.org/neurips/2020/fort2020neurips-stochastic/)

BibTeX

@inproceedings{fort2020neurips-stochastic,
  title     = {{A Stochastic Path Integral Differential EstimatoR  Expectation Maximization Algorithm}},
  author    = {Fort, Gersende and Moulines, Eric and Wai, Hoi-To},
  booktitle = {Neural Information Processing Systems},
  year      = {2020},
  url       = {https://mlanthology.org/neurips/2020/fort2020neurips-stochastic/}
}