English

Reliable training and estimation of variance networks

Machine Learning 2019-11-05 v2 Machine Learning

Abstract

We propose and investigate new complementary methodologies for estimating predictive variance networks in regression neural networks. We derive a locally aware mini-batching scheme that result in sparse robust gradients, and show how to make unbiased weight updates to a variance network. Further, we formulate a heuristic for robustly fitting both the mean and variance networks post hoc. Finally, we take inspiration from posterior Gaussian processes and propose a network architecture with similar extrapolation properties to Gaussian processes. The proposed methodologies are complementary, and improve upon baseline methods individually. Experimentally, we investigate the impact on predictive uncertainty on multiple datasets and tasks ranging from regression, active learning and generative modeling. Experiments consistently show significant improvements in predictive uncertainty estimation over state-of-the-art methods across tasks and datasets.

Keywords

Cite

@article{arxiv.1906.03260,
  title  = {Reliable training and estimation of variance networks},
  author = {Nicki S. Detlefsen and Martin Jørgensen and Søren Hauberg},
  journal= {arXiv preprint arXiv:1906.03260},
  year   = {2019}
}

Comments

Appeared at NeurIPS 2019

R2 v1 2026-06-23T09:47:21.756Z