This project is a CUDA implementation of the paper
Playing Atari with Deep Reinforcement Learning
Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wierstra, Martin Riedmiller (NIPS workshop 2013)
However we have not got the performance reported by the paper yet.
Yuan Gao & Yao Lu
rl-glue, SDL, SDL-image, CUDA, cublas
cd ENV/src/dnq_gpu make cd ../.. make
For example, to play game Breakout
./ale -player_agent deep_q_agent -display_screen true breakout.bin