Inverse Reinforcement Learning (Inverse RL / IRL) is a type of RL where the objective is opposite from forward RL. Instead of learning a policy from a reward function, we are trying to learn a reward function from a policy or demonstration of a task. In this video I go through why to use Inverse Reinforcement Learning, why to use Inverse RL, examples of IRL, some of the theory, and some existing IRL methods. I cover one of the original papers by Andrew Ng, as well as some newer works on Maximum Entropy IRL (MaxEnt IRL), and Adversarial IRL.
RL Theory playlist: https://www.youtube.com/watch?v=1OI0uuz9jkI&list=PL_49VD9KwQ_OML1Knh-Yb7FUFkhTLS0jL
IRL Algorithms paper: https://ai.stanford.edu/~ang/papers/icml00-irl.pdf
MaxEnt IRL paper: https://www.aaai.org/Papers/AAAI/2008/AAAI08-227.pdf
Adversarial IRL paper: https://arxiv.org/abs/1710.11248
