Search Unity

Question Self-play with Hide and seek?

Discussion in 'ML-Agents' started by astrokoala, Jun 2, 2021.

  1. astrokoala

    astrokoala

    Joined:
    Mar 23, 2016
    Posts:
    22
    Correct me if im wrong, self-play is where one team uses inference model while the other team tries to learn, the process then get reversed so that the brain can learn to react to different kind of situation to avoid "over fitting"

    but how to implement this kind of training with 2 teams with different brain? for example in hide and seek game, one brain would be the chaser, the other would be the runner?
     
    astrokoala, Jun 2, 2021
    #1

  2. andrewcoh_unity

    andrewcoh_unity

    Unity Technologies

    Joined:
    Sep 5, 2019
    Posts:
    162
    andrewcoh_unity, Jun 2, 2021
    #2