reinforcement-learning-super-mario-A3C

Learning to play supermario using A3C Reinforcement Learning algorithm
Original A3C paper: https://arxiv.org/pdf/1602.01783.pdf

How to run

First install Super Mario Gym

Then edit settings in main.py and agent.py and simply run

python main.py

Tip

Things to change in the super-mario-bros.lua file to send slightly less data through the pipe:

• Row 48: skip_frames=6
• Row 445: for y=32,206 do

Command to kill all emulators at once

ps | grep fceux | awk '{print $1}' | xargs kill -9