Minimal and Simple Deep Q Learning Implemenation in Keras and Gym
The code itself is 78 lines and is self documenting.
The explanation for the code is covered in the blog article https://keon.io/rl/deep-q-learning-with-keras-and-gym/
I made minor tweaks to this repository.
I made the memory
a deque instead of just a list. This is in order to limit the maximum number of the memory so we can give more weights to the more recent memories.
I also added a load
and save
function for convinience.