Postdoctoral researcher

Updated: over 2 years ago
Job Type: FullTime
Deadline: 17 Aug 2021

The AI and machine learning group is seeking a postdoctoral research fellow in the area of reinforcement learning theory to work on projects related to the performance analysis of entropy-regularized reinforcement learning algorithms.

Such algorithms have received lots of attention lately from the theoretical side of the RL community, although many limitations of the framework remain challenging to address. Most importantly, existing methods have trouble adapting to large state spaces and accommodating forms of function approximation, especially for misspecified models. The aim of this project is to develop better approaches for handling misspecification and design algorithms with suitable performance guarantees under adverse conditions. Besides theoretical guarantees, particular attention will be paid to the practicality of the developed methods, and specifically to their implementability in modern deep reinforcement learning frameworks.

The successful candidate needs to have a strong background in reinforcement learning theory, as demonstrated with papers published on the topic in top machine learning venues (e.g., ICML, NeurIPS, COLT, ALT, AISTATS).



Similar Positions