search

TheButlah / makrl

makrl - modular algorithm kit for reinforcement learning
4 stars 1 forks source link

Implement a Fully Connected Q Network #14

Open TheButlah opened 6 years ago

TheButlah commented 6 years ago

Implement a subclass of ActionModel which is a fully connected network to approximate q function. This will be used as a simple example for any TD methods.

TheButlah commented 6 years ago

OK, I think this is completed, although I can't test it until we get the agent and trainer implemented. I'll mark as closed for now.

TheButlah commented 5 years ago

Reopening due to the fact that we now need this to be n-step and conform with the API