M.A. Zanger, MSc

My current research specializes in epistemic uncertainty quantification, ensemble methods and uncertainty-driven representation learning in deep reinforcement learning. In particular, I focus on analyzing the learning dynamics of deep neural models in this context and how one can influence their ability to quantify epistemic uncertainty in order to develop risk-averse, risk-seeking, exploratory or novelty-avoidant behaviors. I am convinced that uncertainty and risk awareness are paramount to successful deployments of AI and decision-making algorithms.