Expressing uncertainty in neural networks for production systems
Publication date
2021
Document type
Research article
Author
Organisational unit
Scopus ID
Series or journal
Automatisierungstechnik : AT
Periodical volume
69
Periodical issue
3
First page
221
Last page
230
Peer-reviewed
✅
Part of the university bibliography
✅
Abstract
The application of machine learning, especially of trained neural networks, requires a high level of trust in their results. A key to this trust is the network's ability to assess the uncertainty of the computed results. This is a prerequisite for the use of such networks in closed-control loops and in automation systems. This paper describes approaches for enabling neural networks to automatically learn the uncertainties of their results.
Version
Not applicable (or unknown)
Access right on openHSU
Metadata only access