Interactive Imitation Learning in State-Space