Search Unity

New Reinforcement Learning Algorithm

Discussion in 'ML-Agents' started by dani_kal, Apr 21, 2020.

  1. dani_kal

    dani_kal

    Joined:
    Mar 25, 2020
    Posts:
    52
    Hello!!!
    I would like to ask how can I train my agents with my own reinforcement learning algorithm? Especially, I want to use MADDPG algorithm instead of PPO. Do you know which changes I have to do?
    Thank you in advance!!!
     
    dani_kal, Apr 21, 2020
    #1

  2. alvaro_unity903

    alvaro_unity903

    Joined:
    Jan 20, 2020
    Posts:
    21
    Is there any algorithm to implement roadmap? I'm interested in custom algorithms also and it would be nice to know if there is any plan to not overlapping developments with the Unity team.
     
    alvaro_unity903, Apr 23, 2020
    #2
    dani_kal likes this.

  3. ChrissCrass

    ChrissCrass

    Joined:
    Mar 19, 2020
    Posts:
    31
    You can definitely implement your own models using TF2, but it isn't simple to set up. Look at the models.py file and go from there (I can't really help beyond that, but you can look into the custom docker as well). Check the documentation.
     
    ChrissCrass, Apr 24, 2020
    #3
    dani_kal likes this.