Closed jbirkesteen closed 1 year ago
I don't know if I really like this change. It would work fine, so therefore I will approve, but we should maybe consider that it could give confusion as to what these parameters does in other agents, since they are not used by the agents
Hm yeah, I thought a bit about this as well, but decided to go with this approach for 2 reasons:
Lambda
, as alpha
is simply a step size which could be passed to the optimiser for other agents.**kwargs
since they are being passed off the to Player parent class in the __init__
. If you know how to do it and would prefer to keep Lambda
only for the TDAgent, please push an update to this. you can push an update to this. Otherwise merge it. ... and also the gamma means different things for TDAgent and DirectPolicyAgents.
Are you sure that gamma has a different interpretation between the TDAgents and DirectPolicyAgent? I know that it is used for different computations, but it is just a discounting factor in both cases, right, just applied in different ways? Or am I wrong in this interpretation?
Moved the hardcoded hyperparameters from the
__init__
ofTDAgent
to the one inPlayer
such that they can be passed to all Player classes askwargs
.