Data Generation

In supervised learning, we need data generated from the real world.

Self-play allows you to create data by playing against yourself. Used mostly in Reinforcement Learning, MCTS.

Self-Supervised Learning automatically create labels.