Dan Schult

1. The "N. descendants" for the second (uniparse) graph sum to many more than the NetworkX graph. Harder to see for the degree graph, but it looks like the sum...

This is another case where we have to be careful with the definitions! The screenshot from the paper does not tell us whether they are using distance to mean "number...

Hmmmm.... how can the sparse algorithm be faster for the larger network. It only takes half the time. Can we see your timing code?

The function is returning the path (as a list of nodes along the path) not an edgelist (the edges of the path). You can reconstruct the edges by finding the...

I prefer to provide tools for users to build solutions -- rather than many functions for each specific need. The proposed solution only provides a shortest path "edge path" (I'm...

We used the standard formula for the eigenvector centrality. For undirected graphs that is what we compute even though we actually use left eigenvectors to compute it. For directed graphs...

Looking at the paper, I do not see where it says N>=3. It looks like the formulas (and code) work for N=2. So the linked comment I made about N

Thanks for the pointer to chapter 2 about N>=3. And I'm also happy that the wording does not imply that n

I would say: if it doesn't talk about negative weights in the paper and the formulas allow it and the results aren't completely nonsensical, then we should allow it as...

I approved this PR for NetworkX endorsement of SPEC7 understanding that it applies only to the API for NumPy random number generators (as stated in the "Scope" part of the...