Learning fast and efficient hyperparameter control for deep reinforcement learning on small datasets (C04)

• Subject Area: Image and Language Processing, Computer Graphics and Visualisation, Human Computer Interaction, Ubiquitous and Wearable Computing
• Term: since 2023
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 499552394

We aim to develop automated techniques for setting the hyperparameters of deep reinforcement learning (RL) approaches in a sample-efficient manner for small datasets, thus reducing uncertainty. We will create appropriate benchmarks and then investigate the performance of currently existing hyperparameter optimization (HPO) methods, to then develop automated approaches. This will include a dynamically meta-level control with the use of a robust hyperparameter transfer for online HPO for offline RL and neural architecture search.

• DFG Programme: Collaborative Research Centres
• Subproject of: SFB 1597: Small Data
• Applicant Institution: Albert-Ludwigs-Universität Freiburg
• Project Heads: Noor Awad, Ph.D.; Professor Dr. Joschka Bödecker