igraph-help
[Top][All Lists]

## Re: [igraph] knn computation

 From: Szabolcs Horvát Subject: Re: [igraph] knn computation Date: Thu, 25 Jan 2018 14:05:41 +0100

```Hello Filippo,

By "parallel edges" I meant more than one edge connecting the same two
graph vertices. The function works only if this never happens.

Szabolcs

On 25 January 2018 at 09:33, Filippo Santi <address@hidden> wrote:
> Dear Szabolcs,
>
> Than you for the reply. I will check for these issues you mentioned, even if
> I should have created the network without allowing for self loops. Instead,
> I don't have clear what you mean for "parallel edges connecting the same two
> vertices": is it related with the possibility of having more edge attributes
> in the graph or is it related to the multiplexity of the graph?
>
> Filippo
>
> 2018-01-23 21:10 GMT+01:00 Szabolcs Horvát <address@hidden>:
>>
>> Hi Filippo,
>>
>> The message means that your graph has either self-loops (an edge
>> connecting a vertex to itself), or parallel edges connecting the same
>> two vertices. The function does not work on such graphs.  You need to
>> supply a simple graph—take a look at simplify().
>>
>> Also, be aware that knn() does not compute the mean neighbour strength
>> for weighted graphs.  https://github.com/igraph/igraph/issues/987
>>
>> Szabolcs
>>
>> On 23 January 2018 at 19:29, Filippo Santi <address@hidden> wrote:
>> > Dear all,
>> >
>> > I am new to R and even more to igraph. I am working on foreign direct
>> > investment network, and I am currently  trying to compute average
>> > nearest
>> > neighbour degree and average nearest neighbour strength, using the knn
>> > function. My code is as following:
>> >
>> > fdi.graph.2003 <- graph_from_data_frame(fdi.edge.2003, directed = T,
>> > vertices = fdi.attr.2003)
>> > ANND <- knn(fdi.graph.2003)
>> >
>> > However, R returns the following error message
>> >
>> > Error in knn(fdi.graph.2003) :
>> > At structural_properties.c:5889 : Average nearest neighbor degree Works
>> > only
>> > with simple graphs, Invalid value
>> >
>> > Do you have any explanation to try to solve this issue? I tried to
>> > compute
>> > it building the graph from a simple binary edgelist too (the above
>> > message
>> > comes from computations based on a weighted network with plenty of edge
>> > attributes, built from a data.frame) but it is not working either.
>> >
>> >
>> > Thanks for any help,
>> >
>> >
>> > Filippo
>> >
>> >
>> >
>> > _______________________________________________
>> > igraph-help mailing list
>> > https://lists.nongnu.org/mailman/listinfo/igraph-help
>> >
>>
>> _______________________________________________
>> igraph-help mailing list
>> https://lists.nongnu.org/mailman/listinfo/igraph-help
>
>
>
>
> --
> Filippo Santi
> PhD Candidate in Development Economics
>
>
> _______________________________________________
> igraph-help mailing list