What is FinalizeAgent supposed to do and when should it be called?

[Carl Stermann-Lücke]

Hello,

In the tutorial, it says for the AbstractAgent.cs:

FinalizeGame: store the result and update the agent after a game

FinalizeGame: store the final model at the end of the simulation

I suppose that one of the two is FinalizeAgent rather than FinalizeGame. Am I right about that? If so, which one is it? I noticed that FinalizeAgent is never called. What is it supposed to do and where should it be called?

It's also not clear to me what the two descriptions above mean. So assuming the agent learns, than does "update the agent" mean this learning? Is the "final model" everything the agent learned? If so, why should it be stored in the end of the whole simulation?

Yours,

Carl

[dockhorn....@gmail.com]

Hi Carl,

FinalizeAgent was thought to be for storing maybe trained values in the end. Since the user should decide, when to call this function and only the final agent should be transmitted to the competition, we did not included it in the main-processing loop.

A possible use for the method was mentioned in this post:

https://groups.google.com/d/msg/the-hearthstone-ai-competition/muxzJllMRMQ/38ykiKd0BAAJ

Best Regards,

Alex