value_function improvements

jsphon / reinforcement_learning

Python Package For Reinforcement Learning

0 stars 0 forks source link

value_function improvements #46

Closed jsphon closed 6 years ago

jsphon commented 6 years ago

It's a bit ugly because of the way it can handle a list or scalar, which has caused a few calculation issues. But we don't want code that tries to be clever and can handle any situation.

Perhaps we need functions for:

calculating a list/array/1d tensor of states
calculating a scalar
building the graph.