96ea64f3a1aa2fd00c72faacf0cb8ac9-AuthorFeedback.pdf

We will add a series of nu-2 merical experiments to demonstrate the minimax optimality of the model-3 based approach studied herein. Figure 1 depicts some numerical plots of this4 kind. Here, we adopt a (least favorable) example designed in the minimax5 bound in Azar et al. [2] (and also Wainwright [37]), which primarily fo-6 cuses on the effect ofthe discount complexity 11 γ on the sample complex-7 ity. Here, the red text represents the sample size38 barrier,andisremovedinourtheory.