8 Nov
2013
8 Nov
'13
10:57 a.m.
On 11/08/2013 12:53 AM, xenil wrote:
- Is the adjacency matrix normalized when no edge weight is used?
No.
I think that for unweighted katz centrality calculation, using the adjoint matrix, the normalization could be necesary to guarantee convergence.
The convergence is guaranteed by the values of alpha and beta alone. One could force convergence some way or another, but then it would no longer be the Katz centrality, but some other variant.
There is another normalization on the _init.py file before the C++ BGL function call, line 738.
This is the only (optional) normalization left in the code. I removed the normalization from the C++ part. Cheers, Tiago -- Tiago de Paula Peixoto <tiago@skewed.de>