Professor Andreas Krause awarded ERC Consolidator Grant

In the process of digitalisation, learning systems are increasingly supporting decisions in the real world. In his ERC project, Professor Krause seeks to develop new ¡°reinforcement learning¡± methods with a previously unattained level of reliability. The approach combines modern results of machine learning with techniques of robust optimisation and efficient algorithms. Careful modelling of uncertainty makes it possible to act cautiously, but still make better decisions as more data becomes available. These learning methods have various potential practical applications ranging from robotics to medical, industrial and scientific domains.