Publication Type
Conference Paper
Journal Name
The International Conference on Learning Representations
Book Title
ICLR 2021 SimDL Workshop
Publication Date
Page Numbers
1 to 10
Issue
1
Conference Name
Ninth International Conference on Learning Representations (ICLR)
Conference Location
Virtual, Austria
Conference Sponsor
Google
Conference Date
Abstract
This paper considers calculation of prediction intervals (PIs) by neural networks (NNs) for quantifying uncertainty in regression tasks, so as to provide fast, accurate and robust emulators to accelerate scientific simulations. We propose a novel method to learn lower and upper bounds of the PI using independent NNs without defining an exclusive loss. Our method requires no distributional assumption, does not introduce extra hyper-parameters, and can effectively identify out-of-distribution samples and quantify their uncertainty. We demonstrate advantages of our method using a benchmark problem and two real-world scientific applications.
