[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [igraph] Negative weights in betweenness centrality calculation
From: |
Tamás Nepusz |
Subject: |
Re: [igraph] Negative weights in betweenness centrality calculation |
Date: |
Thu, 7 Feb 2013 00:23:11 +0100 |
> I am wondering how the betweenness centrality calculation work on the
> negative weights values.
Chances are that it won't. Betweenness centrality calculates all the shortest
paths in the network (where "shortest" means "having smaller total weight") and
then counts how many times a given edge occurs in shortest paths. If all the
edges are negative, then there is a very high chance that a negative cycle
exists in your graph. By traversing a negative cycle many times, one can
arbitrarily decrease the weight of a path between any two vertices whose path
passes through this negative cycle.
> In our graph, we all have edges with negative
> weights in which larger value means closer connections.
You have to find a transformation that turns similarity scores like yours into
distances without introducing negative values. The "correct" transformation
usually depends on your particular application so there is no single "good" way
of doing it.
--
T.