Efficient Evaluation-Time Uncertainty Estimation by Improved Distillation

Abstract

In this work we aim to obtain computationally-efficient uncertainty estimates with deep networks. For this, we propose a modified knowledge distillation procedure that achieves state-of-the-art uncertainty estimates both for in and out-of-distribution samples. Our contributions include a) demonstrating and adapting to distillation's regularization effect b) proposing a novel target teacher distribution c) a simple augmentation procedure to improve out-of-distribution uncertainty estimates d) shedding light on the distillation procedure through comprehensive set of experiments.

Submitted at the ICML 2019 Workshop on Uncertainty & Robustness in Deep Learning(poster & spotlight talk)

References in corpus (2)

• Distilling the Knowledge in a Neural Network
• Probabilistic Backpropagation for Scalable Learning of Bayesian Neural Networks