Under proper condition, K$-$Nearest$-$Neighbor can be asymptotically optimal for regression and classification.

As sample size n$->$infinity, which choice of k leads to optimal method?

$($choose all answers that are correct$)?$