My research is mainly concerned with Reinforcement Learning (RL). The fokus thereby lies on fundamental research in deep RL.

The Deep Q-Network Algorithm (DQN) as a prominent representative of deep RL algorithms uses (not as the only one) a so called Experience Replay (ER). This can be imagined as a capped List (FiFo buffer) of experienced transistions (state, action, reward, next state). My research is fokused on the ER and possible combinations of it with interpolation techniques. The default ER is filled slowly with real experiences, and I look for possibilities of creating synthetic transitions using interpolation to assist the training.

• Averaging rewards as a first approach towards Interpolated Experience Replay. Wenzel Pilar von Pilchau. In: Draude, C., Lange, M. & Sick, B. (Hrsg.), INFORMATIK 2019: 50 Jahre Gesellschaft für Informatik – Informatik für Gesellschaft (Workshop-Beiträge). Bonn: Gesellschaft für Informatik e.V.. (S. 493-506). DOI:
• Combining Machine Learning with Blockchain: Benefits, Approaches and Challenges. Wenzel Pilar von Pilchau and Jörg Hähner. In Organic Computing - Doctoral Dissertation Colloquium (OC-DDC), 21. - 22.06.2018, Würzburg, Germany