Thank you for liking
You have already liked this page, you can only like it once!
Machine learning (ML) automates tasks that are difficult to program, for example semi-automatically driving rovers. It could be extended to a wider range of tasks if we can better ensure its safety. Currently, the safety of ML components is estimated through traditional methods (e.g., accuracy on a test set) rather than evidence supporting safety arguments (e.g., demonstrating that failing scenarios are rare).
A co-funded research project initiated by ESA Discovery & Preparation is developing a new approach to ML safety that uses clustering algorithms and simulators.