Re: [igraph] How to avoid indexing issues with "which" fuction

[giorgio delzeri]

Ok, I think I wrote a too long email ;-)

So I'll try to simplify my request:

I have a problem with the "wich" and/or "adjacent_vertices" functions.

I think there is a problem about the ids-indexing.

In the part of the algorithm just below:

"

w <- unique(unlist(adjacent_vertices(g, which(V(g)$compartment=="SPREADER"))))

 dummy_1 <- which(V(g)[w]$compartment=="IGNORANT")

  z <- sample(w[dummy_1], ceiling(p*length(dummy_1)))

  V(g)[z]$compartment <- "SPREADER"

  

  dummy_2 <- which(V(g)$compartment=="SPREADER")

  k <- sample(dummy_2, ceiling(r*length(dummy_2)))

  V(g)[k]$compartment <- "STIFLER"
"

it should randommly select one of the adjacent vertices of "SPREADER" vertices that are "IGNORANT" (IGNORANT and SPREADER, such as IMMUNE and STIFLER are all "label" that I assigned to the vertices) and change their label in "SPREADER".

Then, it select a certain percentage of the "SPREADER" and changes their label in "STIFLER".

But there are some indexing issues with that part of the algorithm, because at the end of the process, I found that some of the nodes that were "IMMUNE" at the start of the algorithm (at the beginning of the algorithm, I select a certain percentage of the whole nodes and I call them "IMMUNE", but these nodes have to remain "IMMUNE" during the whole process!!) changes their label in "SPREADER".

So I think the problem is in the fact that these functions "confuse" the ids of vertices with the position in the "adjacent_vertices" vector.

Am I wrong?

How can I solve this problem with the algorithm, do you have any idea or advice?

Thanks in advance.

Giorgio

2017-05-31 22:42 GMT+02:00 giorgio delzeri <address@hidden>:

Hi, I already wrote part of the igraph algorithm which should do this (it is a simulation of a contagion process, but in a positive sense: innovation spreading)

It starts from a scale-free graph.

It randomly select a given percentage (10%) of the whole vertices ad assign them the label (or "compartment") IMMUNE.

The rest of vertices are, instead, IGNORANT.

It randomly select 1 node among the IGNORANT ones, and changes its label in "SPREADER". (so far its all ok with the algorithm, I think).

Now, it starts a while cycle (the cycle stops if there are no more vertices "IGNORANT" and/or there are no more vertices "SPREADER").

It select a given percentage (for example 15%) of the "IGNORANT" nodes that are adjacent to "SPREADER" , and changes their label (compartment) from "IGNORANT" to "SPREADER" (this is the innovation spreading).

It select a given percentage (for example 12%) of the nodes that are "SPREADER" (the whole "SPREADER" compartment, not only the ones that are JUST become "SPREADER") and change their label to "STIFLER" (stiflers are the nodes that were spreader of the innovation, and now they stop spreading).

(END OF THE WHILE CYCLE).

END OF THE ALGORITHM.

nb: note that the "IMMUNE" nodes rest in the same compartment (same label) during the whole process.

Below, the first "edition" of my algorithm:

"
g <- sample_pa(n=100, power=1, m=1, directed= F)

V(g)$compartment <- “IGNORANT”

x <- sample(vcount(g), round(vcount(g)*0.1))

V(g)[x]$compartment <- “IMMUNI”

y <- sample(which(V(g)$compartment==”IGNORANT”), 1)

V(g)[y]$compartment <- “SPREADER”

p <- 0.15

r <- 0.12

while (sum(V(g)$compartment==”SPREADER”) >0) {

w <- unique(unlist(adjacent_vertices(g, which(V(g)$compartment==”SPREADER”))))

z <- sample(which(V(g)[w]$compartment==”IGNORANT”), ceiling(p*sum(V(g)[w]$compartment==”IGNORANT”)))

V(g)[z]$compartment <- “SPREADER”

k <- sample(which(V(g)$compartment==”SPREADER”), ceiling(r*sum(V(g)$compartment==”SPREADER”)))

V(g)[k]$compartment <- “STIFLER”

}.

"

A friend advised me to change from the first version to the version just below:

"

g <- sample_pa(n=100, power=1, m=1, directed= F)

V(g)$compartment <- "IGNORANT"

x <- sample(vcount(g), round(vcount(g)*0.1))

V(g)[x]$compartment <- "IMMUNI"

y <- sample(which(V(g)$compartment=="IGNORANT"), 1)

V(g)[y]$compartment <- "SPREADER"

degree(g,which(V(g)$compartment=="SPREADER"))

p <- 0.15

r <- 0.12

while (sum(V(g)$compartment=="SPREADER") >0) {

  

  w <- unique(unlist(adjacent_vertices(g, which(V(g)$compartment=="SPREADER"))))

 dummy_1 <- which(V(g)[w]$compartment=="IGNORANT")

  z <- sample(w[dummy_1], ceiling(p*length(dummy_1)))

  V(g)[z]$compartment <- "SPREADER"

  

  dummy_2 <- which(V(g)$compartment=="SPREADER")

  k <- sample(dummy_2, ceiling(r*length(dummy_2)))

  V(g)[k]$compartment <- "STIFLER"

  }

"

I have 3 problem I can't solve:

1) The while cycle, and the algorithm, must stops if there are no more nodes in the "IGNORANT" compartment and/or there are no more nodes in the "SPREADER" compartment.

2) I already tried the algorithm, and the number of the nodes that are "IMMUNE", decreases after a certain number of the while cycle. But that vertices must remain in the same compartment during the whole process!

Maybe an issue in the while cycle, i think the problem is the indexing of nodes when it chooses the vertices to change from "IGNORANT" to "SPREADER", and/or the ones to change from "SPREADER" to "STIFLER".

3) I have to revise the algorithm, because I want to obtain, at the and of the process, a print of a matrix that tell me about the number of "IGNORANT", "SPREADER" and "STIFLER" nodes in EACH iteration of the while cycle.

Does anyone know the way to solve these 3 problems?

Thank you in advance!!

Giorgio