home | changes | index | login

July 28 2004

SeeAlso: VisualizingThings.

<Labelleski> I just wrote a section about this for my thesis - it turns out some people have done this in the form of scientific citation networks. Although I've never seen a 3d graph of bibliographic references, some researchers have made 2d graphs.

The problem with a 3d graph is that while all graphs are embeddable in some metric space, you have to use a weighted graph to get a real geometry (i.e. distances) - I know there are ways to calculate weights from citations but I can't remember the paper (It was probably by a guy named Newman)

Microsoft is mapping usenet for some reason - check out netscan.research.microsoft.com/ - the 'tree view' and 'crosspost' things are kinda neat.

The crazy thing about networks is that they're all the same - the statistical distribution of links follows the same rules (called a power-law) for basically any network you can think of: Film actor collaborations, telephone calls, email messages, web pages, word co-occurrences, sexual contacts, electronic circuits, p2p networks, metabolic networks, protein interactions, routers on the internet, co-authorship in scientific journals, food-chain webs, large computer programs like Java Dev. Kit, Linux kernel, MySQL... and lots more.

And even crazier is that the power-law distribution is like fractals - self similarity at all scales. Basically, networks are fractal, which indicates that nature is fractal. Other (non-network) power laws: Internet traffic, price changes in commodities, avalanches, sediment distribution in rocks, branching structure of rivers, changes in rotational velocity of deep-space objects, solar flares, mutations in an organism, extinction, traffic jams.

Uh, but anyway - yeah..those links are pretty cool.

<Brennen> Dude. A more detailed response to this later, but I'm reminded that I need to talk to you more often.

pick a name (required to comment or edit a page)
last edited January 6, 2005