Introduction to our reading group

BeyondRL is an initiative started to bridge the gap between teaching and cutting-edge RL. We believe developments in this field are so promising that it is worthwile to discuss selected topics every now and then. Website and organisation thanks to Tijs Maas and the slides are contributed mainly by our speakers.

Our primary focus is self-supervised model based RL. This paradigm features transparency in the steps taken and by training in a self-supervised setting loss hacking can be avoided. We invite anyone with a background in Self-supervised, Continual and Reinforcement Learning to join. At the same time it is great to see sample efficiency improvements, partial observability & probabilistic decision making.