igraph-help
[Top][All Lists]

## Re: [igraph] how R_igraph_shortest_paths treats 'weights'

 From: Bob Pap Subject: Re: [igraph] how R_igraph_shortest_paths treats 'weights' Date: Wed, 3 Jul 2013 13:29:02 +0200

Hi Tamás,  Thanks for the clarification on the treatment of weights in short distance calculation (the bigger, the further/'lower-value'), and flow-related and community detection algorithms (the bigger the more capacity/similar). This is very helpful. I should also assume that any algorithm using the shortest-paths (e.g. betweenness and/or related algorithms, possibly also in the community-detection bunch) also weighs bigger=lower-value? I may need to recheck carefully my calculations and their interpretation.

Maybe this information could at some stage be inserted in the help pages - I have been perplexed in the different calculations, and would be good to know looking at the documentation. Many thanks!

Bob

Kindly,
Bob Pap

On 3 July 2013 12:10, Tamás Nepusz wrote:
Hi,

Weights are treated as distances during shortest path calculations -- this is what "makes sense" in that domain because you can simply add the distances of individual edges in order to get the distance of the whole path and you cannot do the same thing if weights were considered as similarities. On the other hand, weights are treated as "capacities" in flow-related igraph functions and they are treated as "similarity" scores in community detection.

--
T.

On 3 Jul 2013, at 12:05, Bob Pap <address@hidden> wrote:

> Dear all,
> I have a brief question on the treatment of weights in calculating shortest paths. Are weights considered 'distances' or the opposite - i.e. if the weights are flows and I would consider to vertices close to each other if the weight is large, is shortest path doing the opposite? I would think that weights are treated in the same way throughout, making the question particularly relevant.
>
> Many thanks for solving this small (but important!) point.
> Kindly,
> Bob Pap
> _______________________________________________
> igraph-help mailing list