Repository for reinforcement learning pet projects.
Uses the OpenAI library for implementing a few experimental projects in reinforcement learning, where the agent learns the best strategy given its observation of the environment.
- Coup - implements the card game 'Coup', where the agent learns the best strategy for any state of the game given its current hand, and its observation of the other players' actions. Uses python mesa simulation framework and openAI. Currently in progress, and uses a scaled down version of the game that can be easily extended.
- NoThanks - implements the card game of 'No Thanks' where the agent learns the best strategy for any state of the game. Uses a simplified version of the game at the moment.