You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I agree. Or with boundary matrices $B_k$, just like incidence matrices $I$, but giving the relationships between k and k-1 edges.
The intersection profile is $P = I^T I$. Would this not be something like $A_k = B_k^T B_k$? @Marconurisso any thoughts?
You can get the lower adjacency matrix of order k simply by taking the nonzero elements of the down Laplacian $L_k = (B_k^\top B_k \neq 0)$ and, vice versa, you get the upper adjacency matrix from the nonzero elements of the up Laplacian $U_k = (B_{k+1}B_{k+1}^\top \neq 0)$. The adjacency would then be something like $A_k = L_k \odot (1 - U_k)$ i think.
A collaborator would be interested if we had hyperedge adjacency matrices (see definition in page 10 of this preprint).
Basically, two k-edges are :
The entries of the matrices are 1 for pairs of edges that are * adjacent, and 0 otherwise.
It feels similar to the intersection profile we currently have, but that one only counts the number nodes shared between two edges.
The text was updated successfully, but these errors were encountered: