Feature selection

Hi, is there any need to define "correlation_selection" function and not just use SelectKBest with f_regression?

I guess that correlation_selection allows specifying threshold instead of several selected (k-best) features, but is there any difference above that?