ReinforcementLearning

Reinforcement Learning coursework for MSc

Implementing reinforcement learning for maze solving. The maze is randomly generated each time and a Q-Learning agent is trained for 10 minutes to solve the task.

The following methods are implemented:

1. Deep Q network
2. Dueling Q network
3. Target network
4. Double Q learning
5. Experience replay buffer
6. Prioritised action replay
7. Decaying epsilon greedy exploration