🧵 View Thread
🧵 Thread (16 tweets)
fun fact: there's a technical term for the ~polycule in the top left, it's called the "giant component" https://t.co/sqCOh0BoS9 https://t.co/6XhYCUnPKy
giant components emerge naturally in multiple different random graph models. the idea here would be that the more promiscuous you are the more likely you are to end up banging someone in the giant component. if there were somehow two they'd merge with high probability. tmyk
@QiaochuYuan Slutcloud! https://t.co/IEANOq6ulo
@not_a_hot_girl @QiaochuYuan Tired of huge AWS bills? Try SlutCloud today with a free $100 credit. With turnkey slaanesh-scale media generation and streaming, SlutCloud offers industry-leading performance at a fraction of the price. Best of all, use us however and whenever you want, no strings attached!
@danielbrottman @MichelleAkin i also really do think the math is cool and worth knowing about here. like it's a priori a little unintuitive that in the regime where the giant component appears it's unique and all the other components are *much* smaller. it's cool stuff
@QiaochuYuan @danielbrottman https://t.co/Zp5Pp3eaiU
@QiaochuYuan https://t.co/AOf1um25YX
I think that this is for sparse random graphs. Romance isn't quite random, not sparse, nor restricted to a subset of all humans (except for, say, boarding schools). In real world settings, we need the graph clustering techniques developed by Lescovec et al. which show there is a massive core in the entire graph and that clustering of nodes stops making sense above a certainly threshold of cluster sizes (usually around 1,000 to 10,000 nodes).
