Optimizer.h - Vertex order - Stability

[B4rtDC]

Hi,

first of all, congratulations on this great package! While going over some details of the code, I came across a problem in the segment below :

https://github.com/vtraag/leidenalg/blob/e47a1881721b4e5ca8cf056ea74a884ebee2aaba/src/leidenalg/Optimiser.cpp#L728-L739

If you wan't to realise what comment (731-733) says, shouldn't line 734 be the following?

 if (is_node_stable[u] && partition->membership(u) != max_comm && !is_membership_fixed[u]) 

• Before moving or if you don't move v, membership(v) equals max_comm https://github.com/vtraag/leidenalg/blob/e47a1881721b4e5ca8cf056ea74a884ebee2aaba/src/leidenalg/Optimiser.cpp#L569 https://github.com/vtraag/leidenalg/blob/e47a1881721b4e5ca8cf056ea74a884ebee2aaba/src/leidenalg/Optimiser.cpp#L651 This leads to membership(v) != max_comm being false so the neighbor u won't be placed in the queue.

• If you move the community of v, you call upon the move_node(v, max_comm) function, which changes the assignment of v to max_comm. This also leads to membership(v) != max_comm being false, with the same result.

In both cases, I'm not sure that the node u will ever be placed in the queue as intended. If this is indeed an error, it might also be present in the other functions that work in a similar way.

However, if all of the above is fine, could you explain what is meant with the comment?

Kind regards

[vtraag]

Yes, you are complete right! I already pushed a fix for this in #70. I would still like to run a few tests to quantify the effect on runtime and the resulting optimisation before I merge it. Feel free to take it already for a test!

If you find anything else, do let me know!

[B4rtDC]

Thanks for the quick reply. I missed https://github.com/vtraag/leidenalg/pull/70#issue-627208312 that one.

I'll keep you posted should I come across something else :-)