LSPI(Least Squares Policy Iteration) with TF 1.5.0

LSPI?

Least Sqaures Policy Iteration

Dependency

language

• python3

libraries

• gym
• tensorflow=1.5 (I tried tensorflow 1.13, but it's trapped)
• numpy

Run

python3 main.py

In main.py, you can choose basis function option.

1.gaussian

2.deep_cartpole

3.dan(deep_action_network)_h1

4.dan(deep_action_network)_perd

reference

Batch, Off-policy and Model-Free Apprenticeship Learning