A reinforcement learning agent learning to play osu! using Q-learning. - Work In Progress.
For now, different approaches are in the code but only the most efficient one will be kept in the end.
/!\ Training takes a very long time (~ 2 to 4 weeks) and is not guaranteed to work for now /!\
Osu! is a free rhythm game on computer by Peppy.
ToDo + setup.py
Some parts of the code in this repo are adapted from these GitHub repo:
Rainbow implementation by Kaixhin
DDPG implementation by vy007vikas
Playing Atari with Deep Reinforcement Learning, Mnih et al.
Off-Policy Actor-Critic, Degris et al.
Deterministic Policy Gradient Algorithms, Silver et al.
Continuous control with deep reinforcement learning, Lillicrap et al.
Deep Reinforcement Learning with Double Q-learning, van Hasselt et al.
Prioritized Experience Replay, Schaul et al.
Dueling Network Architectures for Deep Reinforcement Learning, Wang et al.
Asynchronous Methods for Deep Reinforcement Learning, Mnih et al.
Noisy Networks for Exploration, Fortunato et al.
A Distributional Perspective on Reinforcement Learning, Bellemare et al.
Rainbow: Combining Improvements in Deep Reinforcement Learning, Hessel et al.
When to use parametric models in reinforcement learning?, van Hesselt et al.