A simple implementation of Rainbow reinforcement learning to improve the learning performance of DL agents. This project basically follows the idea of the paper Rainbow: Combining Improvements in Deep Reinforcement Learning [arxiv]

The environment is run in python 3.6

To start a reversi game, run:

$ python play.py

In default setting, two A.I agents have been set to play against each other. You may select agents and/or human players to be player1 and player2.

  --interface {cli,gui}
      Interface mode (command-line or GUI).
                        
  --agent1 {human,random,greedy,weighted}
      Agent1 to use.
                        
  --agent2 {human,random,greedy,weighted}
      Agent2 to use.
                        
  --num-episodes NUM_EPISODES
      The number of episodes to run consecutively.

1. A Reversi game engine (completed)
2. Reversi agent A.I. (completed)
3. Deep q-learning (to be done)
4. Rainbow dqn (to be done)
5. Maximum Entropy (to be done)