Best-Arm Identification

Experimental Designs for Heteroskedastic Variance