DDQN : Double Deep Q Network trained for Pong

I implemented DDQN : Double Deep Q Network in Pytorch and trained it to play pong.

After training for 5*10^6 frames, its training became stable at the average score of 19/21 which is near state of the art.

This video is of one the episodes during training.

Link to implementation : https://github.com/chaitanya100100/RL-Algorithms-Pytorch

Видео DDQN : Double Deep Q Network trained for Pong канала Chaitanya Patel