Sinh-arcsinh-normal distributions to add uncertainty to neural network regression tasks: Applications to tropical cyclone intensity forecasts

Barnes Elizabeth A.; Barnes Randal J.; DeMaria, Mark

doi:10.1017/eds.2023.7

i

Sinh-arcsinh-normal distributions to add uncertainty to neural network regression tasks: Applications to tropical cyclone intensity forecasts

2023
By Barnes Elizabeth A. ; Barnes Randal J. ; DeMaria, Mark

Details

Journal Title:

Environmental Data Science
Personal Author:

Barnes Elizabeth A. ; Barnes Randal J. ; DeMaria, Mark
NOAA Program & Office:

CIRA (Cooperative Institute for Research in the Atmosphere)
Description:

A simple method for adding uncertainty to neural network regression tasks in earth science via estimation of a general probability distribution is described. Specifically, we highlight the sinh-arcsinh-normal distributions as particularly well suited for neural network uncertainty estimation. The methodology supports estimation of heteroscedastic, asymmetric uncertainties by a simple modification of the network output and loss function. Method performance is demonstrated by predicting tropical cyclone intensity forecast uncertainty and by comparing two other common methods for neural network uncertainty quantification (i.e., Bayesian neural networks and Monte Carlo dropout). The simple approach described here is intuitive and applicable when no prior exists and one just wishes to parameterize the output and its uncertainty according to some previously defined family of distributions. The authors believe it will become a powerful, go-to method moving forward.
Keywords:

Cyclones Machine Learning Neural Networks (Computer Science) Machine Learning Neural Networks (Computer Science) Sinh-arcsinh-normal Distribution Uncertainty Quantification
Source:

Environmental Data Science. 2023;2:e15
DOI:

https://doi.org/10.1017/eds.2023.7
Format:

PDF
Document Type:

Journal Article
License:

https://creativecommons.org/licenses/by/4.0/
Rights Information:

CC BY
Compliance:

Submitted
Main Document Checksum:

urn:sha-512:f2eb6a99057c5bab172af08def95e7dcde8c938163ff1bcdcf76bdb5a23accf52aca346e0fe4335672e43146b87f9df5c04f8235eb16d678adb6db50eee85da6
Download URL:

https://repository.library.noaa.gov/view/noaa/71976/noaa_71976_DS1.pdf
File Type:

[PDF - 2.49 MB ]

ON THIS PAGE

Details

The NOAA IR serves as an archival repository of NOAA-published products including scientific findings, journal articles, guidelines, recommendations, or other information authored or co-authored by NOAA or funded partners. As a repository, the NOAA IR retains documents in their original published format to ensure public access to scientific information.