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
Perhaps we should have a more general NetSci postmortem, but one specific piece of feedback I got was a request for more graph edit distances. I think we avoided these because we already had Hamming distance, but we should definitely consider a more general GED implementation.
On Jun 4, 2019, at 10:29 AM, Leo Torres ***@***.***> wrote:
Is this the kind of thing that we can implement a single monolithic edit distance and parameterize the hell out of it?
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub, or mute the thread.
Coming around to the idea that we should just not implement this. It would just be a wrapper around NetworkX's graph edit distance, and since this library is already built on top of NetworkX, the value-add is pretty low. The benefit is a consistent interface, but is that worth it? It's not like it's that complicated an interface anyway.
It also opens a whole nother can o worms: does netrd want to implement ALL distances, regardless of whether or not they are accessible through other means?
Activity
leotrs commentedon Jun 4, 2019
First step is to compile a list of edit distances to implement. NetworkX has one, yes?
sdmccabe commentedon Jun 4, 2019
They do. Before the collabathon we had also flagged this survey paper as potentially useful.
leotrs commentedon Jun 4, 2019
Is this the kind of thing that we can implement a single monolithic edit distance and parameterize the hell out of it?
sdmccabe commentedon Jun 4, 2019
sdmccabe commentedon Jan 14, 2020
Coming around to the idea that we should just not implement this. It would just be a wrapper around NetworkX's graph edit distance, and since this library is already built on top of NetworkX, the value-add is pretty low. The benefit is a consistent interface, but is that worth it? It's not like it's that complicated an interface anyway.
leotrs commentedon Jan 14, 2020
Fair. I guess we can close this for now until there is demand?
leotrs commentedon Jan 14, 2020
It also opens a whole nother can o worms: does
netrdwant to implement ALL distances, regardless of whether or not they are accessible through other means?sdmccabe commentedon Jan 14, 2020
I think it's a reasonable goal to have
netrdimplement as many as possible while acknowledging that all is basically impossible.