Funding Agency
RoGeRL develops robust and efficient methods for Automated Reinforcement Learning (AutoRL), enabling reinforcement learning agents to be tuned and configured reliably without relying on tedious, error-prone manual trial and error. The project addresses core AutoRL challenges: hyperparameter optimization, neural architecture search, multi-fidelity optimization and meta-learning, across online, contextual and unsupervised reinforcement learning settings, building on prior work in the field of AutoML. By publishing open-source implementations and benchmarks, RoGeRL aims to establish AutoRL as a reliable standard practice in the RL community, reducing development time and computational cost while improving reproducibility.
Lead at LUHAI: Prof. Lindauer
Funding Program: DFG (German Research Foundation)
Project Period: 36 Month
