You’ll find in this repo the code for three reinforcement learning algorithms :
- LSTD$μ$
- SCIRL
- CSI
the description of which can be found on my research page.
Only SCIRL has a somewhat good, heavily commented implementation. It can be found in tutorials/Exp7.py.
I intend to implement those algorithms properly as a part of some well-known machine learning library. When this is done I will destroy this repo.
In the meantime, feel free to try to make sense of all this. Please don’t hesitate to contact me if you have any question.
Exp1.py : CSI on the inverted pendulum, parameters can be played with.
Exp2.py : Finding out in which areas of the state space of the inverted pendulum the expert is good.
Exp3.py : Trying to find what to plot with CSI on the inverted pendulum.
Exp4.py : Testing different parameters for LSPI and CSI on the Mountain Car.
Exp5.py : Running all algos (SCRIL x2, SCI, Classif, RE) on the Mountain Car.
Exp6.py : Evaluating the policies found in Exp5.
Exp7.py : Running SCIRL on the Mountain Car
Exp8.py : Relative entropy on the mountain car.
Exp9.py : Cascading on the data from Asterix
Exp10.py : Relative Entropy on the Highway
Exp11.py : Plotting the results of different IRL algos on the mountain car
Exp12.py : Running SCIRL on the Highway
Exp13.py : Running CSI on the Highway