This is the RL project of group "State Space Invaders" for the course "Reinforcement Learning" in Fall '18.
The code base for the game partly relies on: this repo. We refactored the code heavily, adapted the environment format of OpenAI Gym and adjusted core mechanics, to adjust to the true state of the game.
Please refer to the folder "notebooks" for the following items:
- "Introduction": Outlining the environment and how to use it.
- "Random Benchmark": Describing an Agent that takes random actions each move.
- "DQN": Training of a DQN model and describing the result.
- "Policy Gradient": Training of a Policy Gradient model and describing the result.