NetLogo: neighbors and eigenvector - networking

Question

I want to count the neighbors and the neighbors neighbors of a turtle to kind of find out a way to calculate the eigenvector-centrality. Since I can´t get the NW:extensions to work and my prof neither. I thought of building a method myself. Since I don´t have that much time until the first round of presentation, I try my best to just count the neighbors and the neighbors neighbors of a turtle.
I decided to have a turtles-own, which counts all neighbors and then sum all the neighbors and neighbors neighbors. But I am stuck in my head and cant get it to work.
set Neighborscount count(link-neighbors)
Anyone has any ideas?
Also a final way to get the eigenvector-centrality to work would be very much appreciated.

Answer 1

First, my main recommendation is to get NW working. I'd be happy to help with that.
Barring that, the below assumes you're working with undirected networks. If that's not the cast, let me know.
The problem with the method that you describe is that it will count some neighbors-of-neighbors multiple times. Instead, you can get the agent-set of neighbors-of-neighbors, and then just count it:
to-report neighbors-of-neighbors
report turtle-set [ link-neighbors ] of link-neighbors
end
Then, [ count neighbors-of-neighbors ] of turtle 0 will give the number of neighbors of neighbors of turtle 0.
Now, a few points:
This will include the turtle itself, since a turtle is always a link neighbor of its link neighbors. If you don't want to include the turtle itself, you can just throw an other in there: [ count other neighbors-of-neighbors ] of turtle 0.
This won't include turtles that are link neighbors of turtle 0, but are not linked to any other neighbors of turtle 0. To add those in, you might consider a procedure like this:
to-report turtles-in-link-radius [ n ]
let result turtle-set self
repeat n [
set result (turtle-set result [ link-neighbors ] of result)
]
report result
end
This can then be used, for instance, like so: [ count turtle-in-link-radius 2 ] of turtle 0 to count all turtles at most 2 hops from turtle 0. This reporter has the same behavior as nw:turtles-in-radius, though it will be significantly slower.
Now, I first noted that your idea would count some turtles multiple times. Given that you're interested in something like eigenvector-centrality, maybe this is what you want. In that case you can do this: [ sum [ count link-neighbors ] of link-neighbors ] of turtle 0 to get it without having to bother with a turtles-own. That said, if you do use a turtles-own, you can actually calculate the eigenvector centrality itself:
turtles-own [
...
next-ec ;; helper variable for iteratively computing eigenvector-centrality
eigenvector-centrality
...
]
to calculate-eigenvector-centralities
ask turtles [
set eigenvector-centrality 1
]
repeat 100 [ ;; the higher this number, the more accurate the value. NW uses 100
ask turtles [
set next-ec sum [ eigenvector-centrality ] of link-neighbors
]
let max-ec max [ next-ec ] of turtles
ask turtles [
set eigenvector-centrality next-ec / max-ec ;; normalize
]
]
end
Call calculate-eigenvector-centralities to calculate all turtles eigenvector centralities. The turtles' eigenvector centralities will then be stored in eigenvector-centrality. Note that this code won't work for disconnected networks: each component has to be normalized independently, which is what NW does, but there's no easy way to do that in NetLogo without using NW.