haoshengzou a86354834c actor critic also works. fix some bugs in nstep_q_return. dqn still trains slow.

2018-03-11 15:07:41 +08:00

..

agent

architecture design patch two

2017-11-06 15:24:34 +08:00

core

working on off-policy test. other parts of dqn_replay is runnable, but performance not tested.

2018-03-08 16:51:12 +08:00

data

actor critic also works. fix some bugs in nstep_q_return. dqn still trains slow.

2018-03-11 15:07:41 +08:00

environment

architecture design patch two

2017-11-06 15:24:34 +08:00

simulator

architecture design patch two

2017-11-06 15:24:34 +08:00

__init__.py

mcts update

2017-11-17 15:09:07 +08:00