Where did the graphology-metrics library go? #501
Comments
|
Everything is here now in a monorepo :) |
|
Great, thank you, found it! |
|
So what is this betweennessCentrality problem then :) ? |
Well still the same problem as before that the higher is the weight of the edge, the less importance it ascribes to the connection. I know this is how it's implemented in NetworkX but I have a feeling it would be great to say it in documentation, because it's counter intuitive. Like when we calculate communities, the higher is the weight, the more connected the nodes are, but then with BC, the higher the weight, the less important that connection is. It's not really consistent in my opinion... What do you think? |
Hum, I haven't thought very much about this before but it does not seem counter-intuitive to me because betweenness centrality is based on shortest paths and when dealing with shortest path higher weights usually mean the cost of traversing an edge is higher. I agree that the perspective is different when dealing with communities or spatial layouts. So to me at least, the conceptual difference is consistent, event if edge weights can mean different things in different contexts (lol graph theory is such a mess). This said I agree that the documentation could be clearer regarding this so that people are not surprised by the outcome. As a workaround, know that the betweennessCentrality metric accept a Also, do you know of other libraries or tools that made the choice the other way around wrt betweenness centrality? It could help me further improve the documentation. |
There is a problem in betweennessCentrality calculation for nodes, but where did the library go to post an issue?
The text was updated successfully, but these errors were encountered: