Some nice graphs of networks on Mark Newman’s academic pages.
I was reading about his work via Roland Piquepaille’s weblog, via boingboing. A nice way of going about seeing clusters in complex and large networks. Start removing the links that link clusters, the long connections. Then the clusters will naturally become visible. Remove links that have little in common with each other. I wonder what the algorithm for that actually is.
I’ve just had a bit of an epiphany. It’s probably not something new, but it’s a new thought to me. I’ve always been a bit more of the reductionist school of thought in the old reduction versus emergence fight. If you understand the elements in a system well enough then the so called emergent properties will become apparent due to the detailed understanding of these elements. Though I’ve always been careful to say that a thorough understanding of these elements includes how they relate to the other elements in the system. That the connections between elements or nodes in a system/network is just as important as the nature of the nodes themselves. If you understood this then you could understand the entire system from first principles.
However I now see that this is not always true. In the case of a network/system that grows with some random factor then you cannot truly predict the emergent principles. So if there is no determinism, which I wholeheartedly disagree with on an emotional level as much as any thing else then you cannot truly predict the result of the system.
Now in most real world cases of this the system does not grow truly randomly. Take the human brain. There must be some degree of predictivity of the growth of that as a system, otherwise the vast majority would be stupid lumps of flesh, and only in exceptional circumstances would the mind occur. So there must be some generally predictable outcome from first principles.
So I don’t think much different than before I now realise after writing this. I’m still in that mode of thought that wants to treat the individual elements as important as well as the links between them, sort of in the same way as light is both a particle and a wave. Emergent properties are important, but as much of the network theory research explosion is proving, that emergent properties are not magic gestalts, but predictable properties of various arrangements of connections.
Sometimes synchronicity just kicks in grabs a hold of some of your current thinking remixes it with a whole other dimension and then takes it to new level.
I’ve been thinking along a whole bunch of lines recently, network theory, cybernetics, cellular automata, complexity, etc. Also I’ve been doing some reworking of my work on reputation, trying to put together a more general theoretical structure. Allowing predictions to be made off it as well as hopefully allowing some openness so that it could even evolve into an even more general theory about social networks. Then I read about a presentation given by Wil Wright of Sims fame about his work on modelling. Yet to read the transcript but this key slide from his talk screams sense. It really brings it together for me.
