It is not clear why the complexity of the algorithm is O(n log n). It could be that during first phase we need to go through all the nodes multiple times until the convergence happens. And this amount could depend on "n". However, if we modify the algorithm so that it has no more than "k" runs for phase 1, the complexity is indeed o(nlog(n))
When you switch slide at around 9:15, there should be a factor 2 in front ok k_(i,in). Because its edges will now be counted twice : once for i, once for the its neighbours in C.
Actually everything seems fine for me here. By definition: k_{i,in} is a double sum of weights between (u,v) where u != v. Because k_{i, in} = \sum_{j \in C}_A_{ij} + \sum_{j \in C} A_{ji} (there was a typo in last term in slides)
On the slide at 12:25 it shows Delta M_0,4 = 0.26. Since nodes 0 and 4 are not connected, I don't think we would compute this change in modularity, correct? We would only compute it for neighbors of node 0?
it is representing the likeliness that a link actually exists between nodes due to a similarity rather than by pure chance. intuitively, looking at the modularity equation, it subtracts the probability of connection between nodes from 1 . This number gives us an idea of how much more likely it is for the edge to exist rather than by pure chance.
![Seungmin "그렇게, 천천히, 우리(As we are)" | [Stray Kids : SKZ-PLAYER]](http://i.ytimg.com/vi/kAzmhLHePqU/mqdefault.jpg)
Thank you for reading the presentation
It is not clear why the complexity of the algorithm is O(n log n). It could be that during first phase we need to go through all the nodes multiple times until the convergence happens. And this amount could depend on "n". However, if we modify the algorithm so that it has no more than "k" runs for phase 1, the complexity is indeed o(nlog(n))
When you switch slide at around 9:15, there should be a factor 2 in front ok k_(i,in). Because its edges will now be counted twice : once for i, once for the its neighbours in C.
Actually everything seems fine for me here. By definition: k_{i,in} is a double sum of weights between (u,v) where u != v. Because k_{i, in} = \sum_{j \in C}_A_{ij} + \sum_{j \in C} A_{ji} (there was a typo in last term in slides)
On the slide at 12:25 it shows Delta M_0,4 = 0.26. Since nodes 0 and 4 are not connected, I don't think we would compute this change in modularity, correct? We would only compute it for neighbors of node 0?
Same doubt. Did you get the answer if yes do share
I'm Louvain it!❤
omg
Thanks, this was excellent!
Is there a keynote shared, Thanks
What exactly modularity is?
Newman, M. E., & Girvan, M. (2004). Finding and evaluating community structure in networks. Physical review E, 69(2), 026113.
Intuitively, I feel like it some how like the standard deviation in statistics. It's a kind of measure how far you leave from the expectation.
@@chrischang1980 Nice observation
Jure explains it in the lecture 13.2 You can check that one.
it is representing the likeliness that a link actually exists between nodes due to a similarity rather than by pure chance. intuitively, looking at the modularity equation, it subtracts the probability of connection between nodes from 1 . This number gives us an idea of how much more likely it is for the edge to exist rather than by pure chance.
Excellent video, but why is he green?