Attention, we start now one hour earlier, at 18:00 o'clock!
In the Paper Discussion Group (PDG) we discuss on a weekly base recent and fundamental papers in the area of machine learning. For several weeks, we follow one track to dive a bit deeper into a topic by reading matching or correlate papers. If you are interested, please read the paper and join us.
We follow the recent success of Deepminds AlphaStar system (https://deepmind.com/blog/alphastar-mastering-real-time-strategy-game-starcraft-ii/) which beats human professional players in the game StarCraft II. Our goal is to understand the techniques behind this success and start therefore an extended track.
Topic: Continuous control with deep reinforcement learning
The potential next papers:
Model-Free Reinforcement Learning with Continuous Action in Practice
Asynchronous Methods for Deep Reinforcement Learning
StarCraft II: A New Challenge for Reinforcement Learning
Attention Is All You Need
IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures
DEEP REINFORCEMENT LEARNING WITH RELATIONAL
A Unified Game-Theoretic Approach to Multiagent Reinforcement Learning
Human-level control through deep reinforcement learning
Counterfactual Multi-Agent Policy Gradients
Population Based Training of Neural Networks
Human-level performance in first-person multiplayer games with population-based deep reinforcement learning
[Wir treffen uns im Informatik-Gebäude des KIT (50.34), Raum -120. Wenn alle Teilnehmer Mutterspachler sind, sind die Diskussionen auf deutsch.]