Matthew Toseland wrote:
> Why not just do what we did? (Construction from Kleinberg)?

Two reasons: first, Kleinberg's model is designed to explain how local 
navigation is possible in small world networks, it's not designed as a 
realistic model of social networks. It has features like constant 
out-degree, Poisson in-degree and an underlying lattice structure that 
clearly aren't realistic for a social network. (That's not to criticise 
the model, because like I said it wasn't designed to be realistic.)

Second, the routing algorithm is based on the assumption that a social 
network can be embedded in one dimension to give (approximately) the 
kind of distribution of edge lengths that Kleinberg's result requires. 
Using Kleinberg's model to generate the network would beg the question. 
To put it another way, we should design the experiment to falsify the 
hypothesis, not to affirm it.

IMHO ;-)

Cheers,
Michael

Reply via email to