A Bayesian Approach to Generative Adversarial Imitation Learning

Y et, one of the fundamental challenges lies in the fact that obtaining trajectory samples from the environment is often very costly, e.g., physical robots situated in real-world.