English

Non-Volatile Memory Accelerated Posterior Estimation

Machine Learning 2022-02-23 v1 Hardware Architecture

Abstract

Bayesian inference allows machine learning models to express uncertainty. Current machine learning models use only a single learnable parameter combination when making predictions, and as a result are highly overconfident when their predictions are wrong. To use more learnable parameter combinations efficiently, these samples must be drawn from the posterior distribution. Unfortunately computing the posterior directly is infeasible, so often researchers approximate it with a well known distribution such as a Gaussian. In this paper, we show that through the use of high-capacity persistent storage, models whose posterior distribution was too big to approximate are now feasible, leading to improved predictions in downstream tasks.

Keywords

Cite

@article{arxiv.2202.10522,
  title  = {Non-Volatile Memory Accelerated Posterior Estimation},
  author = {Andrew Wood and Moshik Hershcovitch and Daniel Waddington and Sarel Cohen and Peter Chin},
  journal= {arXiv preprint arXiv:2202.10522},
  year   = {2022}
}