These deep neural networks are trained by a novel combination of supervised learning from human expert games, and reinforcement learning from games of self-play. David Silver, et al., "Mastering the game of Go with deep neural networks and…