Interpolated-MLPs: Controllable Inductive Bias