Header

UZH-Logo

Maintenance Infos

A General Framework for Uncertainty Estimation in Deep Learning


Loquercio, Antonio; Segu, Mattia; Scaramuzza, Davide (2020). A General Framework for Uncertainty Estimation in Deep Learning. IEEE Robotics and Automation Letters, 5(2):3153-3160.

Abstract

Neural networks predictions are unreliable when the input sample is out of the training distribution or corrupted by noise. Being able to detect such failures automatically is fundamental to integrate deep learning algorithms into robotics. Current approaches for uncertainty estimation of neural networks require changes to the network and optimization process, typically ignore prior knowledge about the data, and tend to make over-simplifying assumptions which underestimate uncertainty. To address these limitations, we propose a novel framework for uncertainty estimation. Based on Bayesian belief networks and Monte-Carlo sampling, our framework not only fully models the different sources of prediction uncertainty, but also incorporates prior data information, e.g. sensor noise. We show theoretically that this gives us the ability to capture uncertainty better than existing methods. In addition, our framework has several desirable properties: (i) it is agnostic to the network architecture and task; (ii) it does not require changes in the optimization process; (iii) it can be applied to already trained architectures. We thoroughly validate the proposed framework through extensive experiments on both computer vision and control tasks, where we outperform previous methods by up to 23% in accuracy. The video available at https://youtu.be/X7n-bRS5vSM shows qualitative results of our experiments. The project's code is available at: https://tinyurl.com/s3nygw7.

Abstract

Neural networks predictions are unreliable when the input sample is out of the training distribution or corrupted by noise. Being able to detect such failures automatically is fundamental to integrate deep learning algorithms into robotics. Current approaches for uncertainty estimation of neural networks require changes to the network and optimization process, typically ignore prior knowledge about the data, and tend to make over-simplifying assumptions which underestimate uncertainty. To address these limitations, we propose a novel framework for uncertainty estimation. Based on Bayesian belief networks and Monte-Carlo sampling, our framework not only fully models the different sources of prediction uncertainty, but also incorporates prior data information, e.g. sensor noise. We show theoretically that this gives us the ability to capture uncertainty better than existing methods. In addition, our framework has several desirable properties: (i) it is agnostic to the network architecture and task; (ii) it does not require changes in the optimization process; (iii) it can be applied to already trained architectures. We thoroughly validate the proposed framework through extensive experiments on both computer vision and control tasks, where we outperform previous methods by up to 23% in accuracy. The video available at https://youtu.be/X7n-bRS5vSM shows qualitative results of our experiments. The project's code is available at: https://tinyurl.com/s3nygw7.

Statistics

Citations

Dimensions.ai Metrics
139 citations in Web of Science®
167 citations in Scopus®
Google Scholar™

Altmetrics

Downloads

476 downloads since deposited on 27 Jan 2021
148 downloads since 12 months
Detailed statistics

Additional indexing

Item Type:Journal Article, refereed, original work
Communities & Collections:03 Faculty of Economics > Department of Informatics
Dewey Decimal Classification:000 Computer science, knowledge & systems
Scopus Subject Areas:Physical Sciences > Control and Systems Engineering
Physical Sciences > Biomedical Engineering
Physical Sciences > Human-Computer Interaction
Physical Sciences > Mechanical Engineering
Physical Sciences > Computer Vision and Pattern Recognition
Physical Sciences > Computer Science Applications
Physical Sciences > Control and Optimization
Physical Sciences > Artificial Intelligence
Scope:Discipline-based scholarship (basic research)
Language:English
Date:2020
Deposited On:27 Jan 2021 08:20
Last Modified:25 May 2024 01:44
Publisher:Institute of Electrical and Electronics Engineers
ISSN:2377-3766
OA Status:Green
Publisher DOI:https://doi.org/10.1109/lra.2020.2974682
Other Identification Number:merlin-id:20306
  • Content: Accepted Version