Not logged in.
Quick Search - Contribution
Contribution Details
Type | Conference or Workshop Paper |
Scope | Learning and pedagogical Research |
Published in Proceedings | Yes |
Title | NOMU: Neural Optimization-based Model Uncertainty |
Organization Unit | |
Authors |
|
Presentation Type | paper |
Item Subtype | Original Work |
Refereed | Yes |
Status | Published in final form |
Language |
|
Page Range | 8708 - 8758 |
Event Title | Proceedings of the 39th International Conference on Machine Learning (ICML'22) |
Event Type | conference |
Event Location | Baltimore, United States of America |
Event Start Date | July 17 - 2022 |
Event End Date | July 23 - 2022 |
Series Name | Proceedings of Machine Learning Research |
Publisher | PMLR |
Abstract Text | We study methods for estimating model uncertainty for neural networks (NNs) in regression. To isolate the effect of model uncertainty, we focus on a noiseless setting with scarce training data. We introduce five important desiderata regarding model uncertainty that any method should satisfy. However, we find that established benchmarks often fail to reliably capture some of these desiderata, even those that are required by Bayesian theory. To address this, we introduce a new approach for capturing model uncertainty for NNs, which we call Neural Optimization-based Model Uncertainty (NOMU). The main idea of NOMU is to design a network architecture consisting of two connected sub-NNs, one for model prediction and one for model uncertainty, and to train it using a carefully-designed loss function. Importantly, our design enforces that NOMU satisfies our five desiderata. Due to its modular architecture, NOMU can provide model uncertainty for any given (previously trained) NN if given access to its training data. We evaluate NOMU in various regressions tasks and noiseless Bayesian optimization (BO) with costly evaluations. In regression, NOMU performs at least as well as state-of-the-art methods. In BO, NOMU even outperforms all considered benchmarks. |
Free access at | Official URL |
Official URL | https://arxiv.org/abs/2102.13640 |
Related URLs | |
Other Identification Number | merlin-id:23344 |
PDF File | Download from ZORA |
Export |
BibTeX
EP3 XML (ZORA) |