Paper presented at SafeAI 2022
SESAME partners Bonn-Rhein-Sieg University and University of Luxembourg have presented a technical paper at the Workshop on Artificial Intelligence Safety (SafeAI) featuring technology advances developed in the SESAME project. The paper titled: Maximum Likelihood Uncertainty Estimation: Robustness to Outliers features benchmarking the robustness of maximum likelihood based uncertainty estimation methods to outliers in training data for regression tasks, and proposes the use of a heavy-tailed distribution (Laplace distribution) to improve the robustness to outliers. In particular, heavy-tailed distribution based maximum likelihood provides better uncertainty estimates, better separation in uncertainty for out-of-distribution data, as well as better detection of adversarial attacks in the presence of outliers.
The Workshop on Artificial Intelligence Safety is held in conjunction with the Association for the Advancement of Artificial Intelligence (AAAI) Conference on Artificial Intelligence and seeks to explore ways to bridge short-term with long-term issues, idealistic with pragmatic solutions, operational with policy issues, and industry with academia, to build, evaluate, deploy, operate and maintain AI-based systems that are demonstrably safe.