Surprisingly structured academic network embedding

[VladimirShitov]

Hi! While I have not used your package myself yet, I can see that it is getting some attention, specifically in the single-cell field. My congratulations :)

Reading the documentation, I was quite surprised by the Google Scholar coauthorship network visualization. I mean this image:

It looks surprisingly structured and almost perfectly symmetrical. I can imagine something like that in mathematical structures, but would you expect to see it in real data? Do you have any explanation for it?

Thank you in advance!

[akshayka]

Hi! I believe we were preserving integral shortest path distances, using an l1-loss, which likely encouraged this "discretized" structure to emerge.