19 — 21 September 2022
Dr. Andrew Thompson
Senior Research Scientist, National Physical Laboratory (NPL)
Machine learning (ML) offers the potential to bring transformation to metrology both by extending its reach and by enabling its automation. Crucial to realising this potential is the ability to evaluate the uncertainty of ML predictions in a principled and scalable way. Case studies in which uncertainty-aware ML has been implemented in metrology applications will be presented. One specific challenge is to propagate data uncertainty through ML models, and a method for doing this in the context of random forests will also be described.