Implement SR and Replay with DNN
Implement Simple and Double DQN on gridworld and catpole
Upgrade Double DQN on discrete and continue space states with the same NN
Linear Successor Feature Neural Network Create a latent space allowing faster transfer learning on reward and transition.