A type of learning used in neural networks that involves learning by trial and error in response to rewards or punishments.