High-Quality Prediction Intervals for Deep Learning: A Distribution-Free, Ensembled Approach

Tim Pearce, Alexandra Brintrup, Mohamed Zaki, Andy Neely
Proceedings of the 35th International Conference on Machine Learning, PMLR 80:4075-4084, 2018.

Abstract

This paper considers the generation of prediction intervals (PIs) by neural networks for quantifying uncertainty in regression tasks. It is axiomatic that high-quality PIs should be as narrow as possible, whilst capturing a specified portion of data. We derive a loss function directly from this axiom that requires no distributional assumption. We show how its form derives from a likelihood principle, that it can be used with gradient descent, and that model uncertainty is accounted for in ensembled form. Benchmark experiments show the method outperforms current state-of-the-art uncertainty quantification methods, reducing average PI width by over 10%.

Cite this Paper


BibTeX
@InProceedings{pmlr-v80-pearce18a, title = {High-Quality Prediction Intervals for Deep Learning: A Distribution-Free, Ensembled Approach}, author = {Pearce, Tim and Brintrup, Alexandra and Zaki, Mohamed and Neely, Andy}, booktitle = {Proceedings of the 35th International Conference on Machine Learning}, pages = {4075--4084}, year = {2018}, editor = {Dy, Jennifer and Krause, Andreas}, volume = {80}, series = {Proceedings of Machine Learning Research}, month = {10--15 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v80/pearce18a/pearce18a.pdf}, url = {https://proceedings.mlr.press/v80/pearce18a.html}, abstract = {This paper considers the generation of prediction intervals (PIs) by neural networks for quantifying uncertainty in regression tasks. It is axiomatic that high-quality PIs should be as narrow as possible, whilst capturing a specified portion of data. We derive a loss function directly from this axiom that requires no distributional assumption. We show how its form derives from a likelihood principle, that it can be used with gradient descent, and that model uncertainty is accounted for in ensembled form. Benchmark experiments show the method outperforms current state-of-the-art uncertainty quantification methods, reducing average PI width by over 10%.} }
Endnote
%0 Conference Paper %T High-Quality Prediction Intervals for Deep Learning: A Distribution-Free, Ensembled Approach %A Tim Pearce %A Alexandra Brintrup %A Mohamed Zaki %A Andy Neely %B Proceedings of the 35th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2018 %E Jennifer Dy %E Andreas Krause %F pmlr-v80-pearce18a %I PMLR %P 4075--4084 %U https://proceedings.mlr.press/v80/pearce18a.html %V 80 %X This paper considers the generation of prediction intervals (PIs) by neural networks for quantifying uncertainty in regression tasks. It is axiomatic that high-quality PIs should be as narrow as possible, whilst capturing a specified portion of data. We derive a loss function directly from this axiom that requires no distributional assumption. We show how its form derives from a likelihood principle, that it can be used with gradient descent, and that model uncertainty is accounted for in ensembled form. Benchmark experiments show the method outperforms current state-of-the-art uncertainty quantification methods, reducing average PI width by over 10%.
APA
Pearce, T., Brintrup, A., Zaki, M. & Neely, A.. (2018). High-Quality Prediction Intervals for Deep Learning: A Distribution-Free, Ensembled Approach. Proceedings of the 35th International Conference on Machine Learning, in Proceedings of Machine Learning Research 80:4075-4084 Available from https://proceedings.mlr.press/v80/pearce18a.html.

Related Material