search

geek-ai / MAgent

A Platform for Many-Agent Reinforcement Learning
MIT License
1.68k stars 332 forks source link

how to use 'parameter-sharing DQN' in my own task? #46

Closed OwalnutO closed 5 years ago

OwalnutO commented 5 years ago

Thanks for your brilliant code.

I want to know how can I use the RL algorithm in my onw task? For example , If I have my own environment to provide reward and many agents, how can I use your RL code to train on my dataset? In other words, I just want to know what is 'parameter-sharing DQN' and which part of the code I can use to train my own dataset?

Kipsora commented 5 years ago

Thank you for being interested in our project.

The parameter-sharing DQN is targeting on optimizing the Environment defined here, you can create your own class which inherits from it. However, the builtin model in MAgent is specialized so it is not recommended to directly apply the builtin model to your own environment. A better way to do this is to just read through these codes and implement yourself.