Origin of "Bad" Hub

The experiment indicates that high dimensionality and skewness of $N_k$ are not sufficient to induce "bad" hubs. Instead, based on the former fact, we can argue that there are two, mostly independent, forces at work:

• violation of the cluster assumption
• high intrinsic dimensionality

"Bad" hubs originate from putting the two together; that is, the consequences of violating the cluster assumption can be more severe in high dimensions than in low dimensions, not in terms of the total amount of "bad" k-occurrences, but in terms of their distribution, since strong regular hubs are now more prone to "pick up" bad k-occurrences than non-hub points.