I Want to use DQN

Hello, I am an undergraduate research student who is using Unity ML Agent to study reinforcement learning.

I need to use DQN in ML Agent.
Therefore, I searched a lot on Google.
However, Google's results show that we are having a hard time designing the algorithm because we do not have the latest version of the code.
Is there any code that can use DQN in the latest version?
If not, is there a way to create a Unity environment in the latest version, build it, and run it through Python?

I need to use DQN in ML Agent.
Therefore, I searched a lot on Google.
However, Google's results show that we are having a hard time designing the algorithm because we do not have the latest version of the code.
Is there any code that can use DQN in the latest version?
If not, is there a way to create a Unity environment in the latest version, build it, and run it through Python?
I tried to learn using Python using the documentation on github, but there is a problem that the learning does not proceed, so please contact us.

I Found a Configuration File.

network_settings:
normalize: false
hidden_units: 256
num_layers: 1
vis_encode_type: simple

Is it right to be able to use DQN with this?

There is a google colab implementation of dqn. https://colab.research.google.com/drive/1nkOztXzU91MHEbuQ1T9GnynYdL_LRsHG#forceEdit=true&sandboxMode=true&scrollTo=KkzBoRJCb18t
Also you can use the gym wrapper and use dqn from other repos

1 Like