DQN

The DQN algorithm used for solving Gym's cartpole environment.
I changed the reward function:

reward = np.cos(2*next_state[3]) 

Requirements

• Tensorflow
• Numpy
• Gym

Run

There is a constant:

DEVICE = '/gpu:0'

Set it to '/cpu:0' if you don't have one.

And then run as:

$ python main.py