On community outliers and their efficient detection in information networks