Skip to content

darknovismc/Reinforcement-Learning-Breakout

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 

Repository files navigation

Reinforcement-Learning-Breakout

Simple breakout game with DQN agent which learn how to play it. Written in C++17 in Microsoft Visual Studio 2019. The game itself is written in OpenGL using freeglut library. Standard DQN algorithm consisting of 2 neuro-network models with replay memory learns from game's ram data using epsilon-greedy policy. Each model is a dense layer with 5 input states(2xball positions,paddle position,2xball velocities) and 3 output actions (left,right,none). I used version 2.8 tensorflow C api so please include missing dll and replace the header files if needed. You need to train the agent for 1 to 2 hours to achieve good game results.