lossv/yueyue_2048_dqn

play the 2048 by DQN

yueyue_2048_DQN

Play the 2048 with RL Nerul Net Work, implement the DQN and mc serch algorithm

Although I adjust the weights and some parameters, but it's only achive 1024.

So I use a Montcarl to play this game,but it's only achive 2048 can't surpass 2048 achieve 4096 or bigger.

How to play?

run the main.py to train the network and you can also change the param in the utils.py befor run main.py

You can run ai_algo.py to play ,it can reach 2048 almost every times, and sometimes can reach 4096

reference [1].ε–MDPs: Learning in Varying Environments