English

Amortized variance reduction for doubly stochastic objectives

Machine Learning 2020-03-26 v1 Machine Learning Methodology

Abstract

Approximate inference in complex probabilistic models such as deep Gaussian processes requires the optimisation of doubly stochastic objective functions. These objectives incorporate randomness both from mini-batch subsampling of the data and from Monte Carlo estimation of expectations. If the gradient variance is high, the stochastic optimisation problem becomes difficult with a slow rate of convergence. Control variates can be used to reduce the variance, but past approaches do not take into account how mini-batch stochasticity affects sampling stochasticity, resulting in sub-optimal variance reduction. We propose a new approach in which we use a recognition network to cheaply approximate the optimal control variate for each mini-batch, with no additional model gradient computations. We illustrate the properties of this proposal and test its performance on logistic regression and deep Gaussian processes.

Keywords

Cite

@article{arxiv.2003.04125,
  title  = {Amortized variance reduction for doubly stochastic objectives},
  author = {Ayman Boustati and Sattar Vakili and James Hensman and ST John},
  journal= {arXiv preprint arXiv:2003.04125},
  year   = {2020}
}
R2 v1 2026-06-23T14:08:45.265Z