How can I train a neural network with previous games? [closed]

Suppose I want to teach a neural network to play chess. I did the reinforcement learning method. I've seen cases where these networks learn from game examples, but how can this be done?