Google DeepMind Artari Plays Pacman
PhyXs Vision PhyXs Vision
100K subscribers
95,562 views
0

 Published On Aug 2, 2015

Hi, I am in the process of training the DQN to learn playing Pacman,this screen cast was taken when at 5 million steps. It's not satisfying any way, but we can see a bit from the video that the agent is really learning. I will let the training continue, it may finally surprise us at some time.
I use Geforce 760(4G version), it can get 3-4 million steps each day, if I could afford a nicer GPU, it's can be done in couple of days.

show more

Share/Embed