Reliable training and estimation of variance networks

Nicki Skafte Detlefsen, Martin Jørgensen, Soren Hauberg

Abstract: We propose and investigate new complementary methodologies for estimating the predictive uncertainty of 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 . Our experiments shows significantly improved predictive uncertainty estimation over state-of-the-art methods.

Code Link: https://github.com/SkafteNicki/john

CMT Num: 3425