Reinforcement learning experiments

Learning RL by implementing and analysing different RL methods from scratch.

Directory	Game	Number of agents	RL method
nim-dqn	Nim-21	2	Deep Q-network
nim-a2c	Nim-21	2	Advantage Actor Critic
matching-pennies-a2c	Matching Pennies	2	Advantage Actor Critic
snake-a2c	Snake	1	Advantage Actor Critic
snake-ppo	Snake	1	Proximal Policy Optimisation

MLFlow

I'm also using this project to learn more about MLFlow. Some of the train scripts depend on an actively running tracking server. Please check MLFlow documentation on how to start a tracking server and set the MLFLOW_URI environment variable to the correct tracking server URL.