We currently have an open call for postdoctoral researchers within continual learning and learning with dataset shift. There are two direction for this position and we are looking to recruited a postdoctoral research for both or either of the two directions.

Direction 1 entails fundamental research on continual learning for dataset shift with data efficient and interpretable models and algorithms. This direction investigates methods for learning, updating, and forgetting for domains where data is incrementally available or changes over time (e.g. models of system dynamics, robotic mapping, reinforcement learning, or environmental data). Prior knowledge, experience, and interest in probabilistic and/or statistical machine learning is expected for this direction. In addition, prior knowledge, experience, and interest in Bayesian methods and Gaussian processes are advantageous.

Direction 2 entails fundamental research on reinforcement learning with continual adaptation and domain shift including, but not limited to, meta and transfer learning methods. This direction investigates methods for fast and efficient reinforcement learning in real-world settings by exploiting knowledge transfer, e.g. from other domains or experience, to allow faster learning and adaptation in the target environment. Prior knowledge, experience, and interest in reinforcement learning in continuous spaces and deep learning are expected for this position. Prior knowledge, experience, and interest in robotic perception and control as well as probabilistic machine learning are advantageous.