If you'd like to support these videos:
  / notdavid  

This is definitly the most effort and mathematics I've ever put into calling someone a nerd. I hope you enjoyed it! I also didn't know about this until it was pointed out to me, but https://youtubeatlas.com has a very similar network to mine. Our networks were made differently but conceptually are similar. Highly recommend you check it out!

Made using #blender as well as Davinci Resolve and Gephi.


DISCLAIMER: My channel is not endorsed by nor associated with any individuals, creators, or organizations featured in this video.
But also I hope that anyone that was included interprates their inclusion as a sign of my appriciation for their content.

Creators included (whose names were not explicitly written on screen):
Jenny Nicholson, Summoning Salt, Matthewmatosis, Technology Connections, Drawfee Show, Super Eyepatch Wolf, Jacob Geller, Hbomberguy, Folding Ideas, Legal Eagle, Polygon, Brian David Gilbert, Karolina Żebrowska, Bernadette Banner.

Content Featured:
The qoute at the end was taken from Matthewmatosis:    • Meta Microvideos  
The speedruning video was featuring Darbian, but it was from the Summoning Salt channel:
   • Mario: The Infamous History of Level 5-2  


Chapters:
0:00 I need a hobby lol
1:38 Building the network
4:26 Who do you subscribe to?
6:46 co-subscribers and bots
9:29 Proving you're nerds
11:37 The Youtube network
14:20 The Youtube communities

Music (in order):
RPG Store - Chris Doerksen https://chrisdoerksen.bandcamp.com/al...
Native - HOME https://open.spotify.com/artist/2exeb...
Killer Vacation - Chris Doerksen (see previous link)
Literally Chill - Florida Skyline https://floridaskyline.bandcamp.com
Intermission - Stux.io and Vaporwavez https://stuxio.bandcamp.com/album/qua...
Vague Familiarity - Not David    • Vague Familiarity  
Atlas - HOME (see previous link)
Power Cycle - Kuraine/ Lena Raine https://radicaldreamland.bandcamp.com...


Notes:
-Regarding the degree distribution and cutting it off abruptly - in actuality the fact that it is a straight line in log-log is really interesting, but that could be a video on its own so rather than going into it, I just decided to leave it as a joke because this is actually one of the things you would do first in a real network analysis.

-I didn't specify this but when I removed the weakest links from the co-citation network, I also removed any nodes that became disconnected from the network.

-In order to make the youtube network I had to play a bit of a trick. You see, the API does not (as far as I know) let me just randomly pick a user from youtube. Because of this, what I did was to get a random list of my subscribers. Then I get a random list of people that they subscribe to. Then I ask who do these people subscribe to, and that is how I build the network. Essentially I'm adding extra steps to the original method to "get away" from my subscribers. Because of this, it is also possible to get channels that do not subscribe to anyone, leading to an isolated node - these were discarded from the analysis.

- For the community detection the Louvain algorithm returns something like 140 communties. However, many of these are very small and while I didn't do additional testing, I doubt they would be statistically significant. These channels were removed from the animation so that I was left with only something like 8 fairly large communities. They were included in all the other analysis however, they just aren't rendered.

- I've put the co-subscriber network gephi file on github
https://github.com/notDavidsGit/notda...