Imitation Learning

Learning a policy with RL can be very inefficient and require a lot of trials. An alternative is to learn directly from expert data.

This approach is referred to as “imitation learning” (IL).