Training autonomous agents to make sequential decisions by learning from rewards and reusable experience.