From the course: Reinforcement Learning Foundations

Inverse reinforcement learning

From the course: Reinforcement Learning Foundations

Inverse reinforcement learning

- [Instructor] Imitation learning is another type of reinforcement learning that you should know about. One of the many ways it's achieved, is through inverse reinforcement learning. As the name portray, an agent tries to imitate the actions of humans directly. A good example of this is in a self-driving scenario, that has both human drivers and reinforcement learning agent drivers. If these reinforcement learning agents use imitation learning to learn from the human drivers, they observe the behavior of the human driver, and learns his objective, values, and rewards. They also learn what the goal of the human driver is, and how they can get there. The details of this method and authors mentioned in this section are above the scope of this course. Definitely read more about them. I would recommend for imitation learning, "The Survey of Inverse Reinforcement Learning" paper, which covers the challenges, methods, and progress of inverse reinforcement learning.

Contents