This Repository holds: A python implementation of 2048 in Numpy for fast computation
A Monte Carlo roll out model that computes smooth action values, where a suboptimal move is just discounted through its negative influence on the mean.
To start and view the agents just run the python scripts
-
montecarlo.py (requirement numpy)
-
qvaluenetwork.py (requirement numpy, keras, tensorflow as backend)
-
a2c.py (requirement numpy, keras, tensorflow as backend)