Skip to content

Implementation of a Reinforcement Learning algorithm for the game Slither.io using the Universe framework from openAI to emulate the game.

Notifications You must be signed in to change notification settings

Bifrost-Research/RL-Universe

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

36 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

What is in this repo?

Implementation of Reinforcement Learning algorithms in Python3 using he OpenAI framework and TensorFlow as neural network libraries. Results are shown with TensorBoard.

#Gym-A3C Implementation of the A3C algorithm for OpenAI Gym's Atari games. The implementation uses processes instead of threads to achieve real concurrency.

##How to run the A3C ? To launch the A3C with the default parameters, just use the following command. It is possible to see the available hyper parameters with the command -h.

python3.5 main.py Pong-v0

To see different plots like the rewards, the losses... it is necessary to launch TensorBoard with the command

tensorboard --logdir=tf_logs/

##Results

Pong-v0, 4 actors, 5 local steps and updating the network with an Adam Optimizer (learning rate: 1e-4).

#Useful papers A3C

About

Implementation of a Reinforcement Learning algorithm for the game Slither.io using the Universe framework from openAI to emulate the game.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages