שאלה | 4. [Imitation Learning] Which of the following statements is true regarding Dagger (multiple answers may apply):

4. [Imitation Learning] Which of the following statements is true regarding Dagger (multiple answers may apply):

		a) DAgger is likely to perform more poorly in large state spaces with many infrequently-visited states.
		b) DAgger offer a good trade-off of being computationally-efficient and having a low regret bound compared to other imitation learning approaches.
		c) The feedback provided by the human expert is more effective at the beginning of the trajectory, as it enables earlier corrections and the prevention of cascading mistakes.
		d) The algorithm is less likely to perform well when there are large differences between the actions chosen by the model and the feedback of the human expert.