search

miyosuda / async_deep_reinforce

Asynchronous Methods for Deep Reinforcement Learning
Apache License 2.0
592 stars 192 forks source link

Large action space #3

Open gowthamnatarajan opened 8 years ago

gowthamnatarajan commented 8 years ago

Is this implementation suitable for Large action spaces? Like 1000s of possible actions? Also, can this be modified to make it work with the problem I am trying to solve? I would like to use this implementation for modify the state and action spaces.

miyosuda commented 8 years ago

@gowthamnatarajan Sorry I'm not sure this implementation can be used for large action space. (I tried to start toy program to check with 1000 actions, but couldn't check well.)

I'm now testing OpenAI gym environment in "gym" branch, so if you want to try another problem, please change game_state.py and ACTION_SIZE in constants.py in gym branch.

gowthamnatarajan commented 8 years ago

I have about 1500 actions. Will checkout your other implementation. Google has a paper for handling large action space https://arxiv.org/pdf/1512.07679.pdf