search

mlii / mfrl

Mean Field Multi-Agent Reinforcement Learning
MIT License
369 stars 101 forks source link

How to get average action during test？ #14

Open GuiyangLuo opened 5 years ago

GuiyangLuo commented 5 years ago

Hi, recently I am trying to reproduce your work and feel a little confused about following question：

  1. How to get average action during test？ During training， we can obtain average action. However, during test, such action cannot be observed. I have carefully read your code, and found that you have adopted the average action in last step?
  2. How you calculate the average action among neighbors? In your implementation, it seems that each node can observe all other players, thus the average is taken over all players? It's really strange.