Imitation Learning from Observations by

Minimizing Inverse Dynamics Disagreement