Alexander van der Grinten
Alexander van der Grinten
It would be good to some simple benchmarks to check if the performance increased substantially. Also, it would be nice to support COO arrays directly (in case the user does...
Well, it's just a matter of adding a function that access 3 numpy arrays instead of one scimy matrix. Once that is added, the scipy-based function can call this new...
This issue has been reported before. One possibility is guaranteeing that the edge ID ordering matches the iterEdges() ordering, but that is not done by the current implementation.
We probably want to defer this PR until the next release is out. There are too many open questions to merge it now.
We should not try to support multigraphs.
If only few of the "non-central" vertices are connected, you can save a lot of work by not computing all pair-wise distances since they will be symmetric anyway. (If the...
I'm not sure whether preventing multi-edges should not be the default. Multi-edges are not supported by NetworKit anyway. The only issue is that this might add a non-trivial runtime cost...
Alright. I think it would be fine to merge the PR as-is. If we later find that these generators still need to be adjusted, we can do it in a...
Pinging @manpen who wrote the code. One question is if we should revert the behavior, or keep the current one and add some big warnings to the release notes.
The mean degree in the undirected / directed cases only differs because more edges are allowed to exist, right? In other words, the average degree in the undirected case is...