Deep RL for Hanabi

Hanabi is a cooperative card game with imperfect information. This multi-agent partial observability makes it an interesting challenge for deep RL algorithms, being called “a new frontier for AI research” in a paper by DeepMind. On 9 September 2021 I finished this master’s research project by presenting it in the Blauwe Zaal at the TU Eindhoven. See the presentation below.

Read my thesis here.

The final presentation on 9 sept 2021