May 24, 2004 7:00 PM   Subscribe

Modelling err.. something.
posted by Gyan (11 comments total)
Exciting stuff. Go small-world networks!
posted by jeffj at 7:35 PM on May 24, 2004

Small world networks are pretty cool right now. Unfortunately I cannot seem to find anything else about this project. However, it was just a matter of time before someone applied the "small world" or "scale free" network formalism to neural networks. Actually, searching the recent literature for "small world neural network" shows a few hits. This makes me wonder what makes this study unique. Interesting stuff, though, Gyan.

Here is a short description of so-called small world networks. Also, a collection of links about small world and scale free networks.
posted by jjray at 8:00 PM on May 24, 2004

posted by solistrato at 9:01 PM on May 24, 2004

Way cool.
posted by spazzm at 9:08 PM on May 24, 2004

We're slaves to machines operateing on machine (clock) time ever since the start of the industrial revolution.

In a fractal sort of way it would be interesting the external networks people build socially mirrors the internal network structure of the brain. It hints at some greater universal order of things.
posted by stbalbach at 9:12 PM on May 24, 2004

Thanks for the links, jjray

stalbach, for the crimes of:
1. Using the term "We're slaves to machines"
2. Using the term "a fractal sort of way"
you have been sentenced to termination.
A cadre of soul-less killing machines will be by to pick you up shortly, please stay put. BOOO!
posted by spazzm at 9:30 PM on May 24, 2004

Via jjray's link, this New York Times contained an analogy that helped an unmathematical person visualize what they mean when they talk about a small world network:

Among people, the short cuts are well-connected individuals -- those who know many people from other countries, different disciplines or unusual subcultures. Such highly connected people establish surprising interconnections between groups that seem impossibly distant from one another, like Iowa farmers and Kalahari bushmen. For example, sometimes a farmer's son will join the Peace Corps in Africa and thus unexpected links are born.
posted by Hildago at 9:37 PM on May 24, 2004

Testing: can one post a comment from (the new improved, apache-served, MeFi)? Sorry, Gyan.
posted by gleuschk at 5:53 AM on May 25, 2004

Processing spazzm's directive... presuming "a fractal sort of way" is a reference to Mandelbrot's "fractional dimensionality". Referenced sentence still makes very little sense. Killbots dispatched.

The neat trick is in getting self-sustaining operation in a simple network with several available stable states using a mechanism that could feasibly evolve without somebody poking it with a stick.

The solution that they're using (looping back the excitatory links in that way) means that John White's "rigorous mathematical description" will have to deal with self referential functions... where a formal description can require non-linear differential equations... which would probably have biomedical engineers like Dr. White waking up screaming in the middle of the night.
posted by snarfodox at 8:16 AM on May 25, 2004

Hildago, I think the article is about something different than the social networks you mention.

The basic issue here is that researchers have been able to model the behaviour of neurons with a very simple mechanism. The behaviour is persistent 2 statefulness or 'bistability'. The characteristic of the networks which makes this happen and also makes them 'small world' networks turns out to be internal connections, not external connections.

The way I understood it was that the networks are internally connected: a to b, b to c, and c to a. The internal connections are what give the persistent state, and also allow the states to be switched.
posted by daver at 8:52 AM on May 25, 2004

post-preview: Doh, I think snarfdox nailed it...
posted by daver at 9:26 AM on May 25, 2004

« Older Threat of famine in Sudan   |   Die Duck Die Newer »

This thread has been archived and is closed to new comments