based on: evangravelle's work
https://gist.github.com/evangravelle/56defd6d01ee738f7cbeb84013145a43
env: openai gym 'MountainCar-v0'
observation:
[0] position [-1.2,0.6]
[1] velocity [-0.07,0.07]
actions:
[0] push left
[1] no push
[2] push right
setting:
features: 16*16 RBF
episode length: 500 per trial
step length: 2000 per episode
hyperparameters:
alpha=0.1
lambda=0.5
gamma=0.99
epsilon0=0.1
epsilonf=0.01
results:
learning curve:
final Q:
