Visual Adversarial Imitation Learning using Variational Models

Reward function specification, which requires considerable human effort and iteration, remains a major impediment for learning behaviors through deep reinforcement learning.