Reinforcement learning is a type of machine learning that allows algorithms to learn by trial and error. By providing positive reinforcement when the algorithm makes a correct prediction, and negative reinforcement when the algorithm makes a mistake, the algorithm can learn to improve its predictions over time.
Some of the emerging trends in reinforcement learning include the use of deep learning for improved function approximation, the use of off-policy methods for improved sample efficiency, and the use of transfer learning to adapt learning to new environments.