A Divergence Minimization Perspective on Imitation Learning Methods

The left column presents results obtained from trained policies. The right column presents results obtained from a random policy.