Self-Assembling Networks

prostoalex writes "Researchers from Humboldt University found a way to build self-assembling networks. By emulating the behavior of ants and insects the team, which is led by Frank Schweitzer, demonstrated a simulation where agent-based architecture was able to quickly assemble itself into a network and quickly react to a broken link or damages. Schweitzer's research papers are available off his personal Web site. The scientific paper referred in the original article, Self-Assembling of Networks in an Agent-Based Model is available off Cornell server."

Comments

We already have one of those.

[Maradine]

My network team looks *just* like a swarm of ants when the network goes down.

Suddenly the expression...

[routerwhore]

"this network looks like a bunch of spiders having an orgy" has new meaning...

Self-assembling intelligence next?

[wiggys]

Taking this idea one step further, what if each computer node on the network was given a basic set of rules so that it emulated a bunch of brain cells. Would the network self-organise to create some sort of intelligence?

Re:Self-assembling intelligence next?

[Neuronerd]

It turns out that already today all successful applications of socalled "artificial intelligence" are self assembling.

In the first approaches to artificial intelligence people used programming languages to obtain systems that generate intelligent or at least apparently intelligent behavior.

All newer approaches to artificial intelligence start with a large number of very simple units that, learning from data from the real world, develop specific patterns of connections. Many models even develop their own structure in such a way.

From my perspective is intelligence as well as artificial intelligence only possible in a system that can self-structure.

Re:Self-assembling intelligence next?

[Alea]

This would be patent nonsense, if the statement itself had any real meaning. First of all, what is meant by "artificial intelligence", "successful", and "application", in this context.

And what does "self-assembling" or "self-organizing" mean, really? The utexas link is pointing to a bunch of machine learning stuff (I research and publish in AI, sometimes in machine learning) that is frankly quite out of date (no kernel machines, SVMs, or any recent clustering techniques). Unsupervised learning can be seen as some sort of self-organization, but it's certainly not the basis of "all successful applications" of A.I.

The claim that "a number of simple units" being organized into some structure is somehow self-organizing is just plain bizarre. Your computer's memory is a big collection of bits being organized by the programs run. The neural nets, Bayes nets, self-organizing maps, etc. listed on that link are not independent agents communicating to form a structure. They are variables in a program, plain and simple, with one big algorithm massaging them into something useful. True, they might exploit local relationships between certain members (e.g. Bayes nets) but so do many algorithms. You might just as well call QuickSort self-organizing.

Depending on your definition, self-organizing computer systems are either so common as to be uninteresting, or so rare that we pretty much never see them in practical applications.

I know it it's "cool" to see this stuff as some sort of biological meta-machine, but to suggest that this is the only useful viewpoint, or even the dominant one, is simply ridiculous.

centralized? self asembly?

[IAR80]

For now I stick to OSPF. And it is not centralized also. And so are BGP, RIP an ISIS.

Maybe we should use the borg icon for this one...

[MeanE]

"We are your network...ect..ect...we will adapt"

this isn't news

[potaz]

Last Update: 18 March 1999
The article was posted to his web site in 1999 and this is front-page stuff? And the article itself was published in 1997. Stop the presses!

I can see it now!

[Neck_of_the_Woods]

HUB, "MALFORMED PACKET!!!! AHHH!!!! - HELP HELP HELP! I am lost!"

Router "Calm down, this is nothing compared to the broadcast storm of 93. Everything will be alright."

HUB, "Thank you,"

Router "These simpletons, when will they ever learn just to ignore that packet."

ala - bugs life.

Ant-like-technology

[LamerBunny]

"By emulating the behavior of ants and insects..."

(I wonder who played the Queen...)

I didn't know ants were this advanced! This must be the final proof that indeed insects are super-intelligent aliens come to earth to eat our... ehm... sugar-water... If only we can harness this power elsewhere! Maybe we should try milipede power-plants next... All that static electricity from all those legs must be harnessed!

Potential

[perspex]

This could be really cool for ad-hoc wireless networks.

these guys don't watch enough anime

[andih8u]

Anyone who does knows this is just a step before the evil computer AI infecting all of the other computers in the world and setting about to destroy mankind. I will rise up to defeat this terrible menace right after I find a girl with blue hair and eyes the size of dinner plates.

A good research work

[varjag]

The paper is indeed very interesting and innovative, but keep in mind that it is very far from being suitable to embed into your next 802.11 adapter.

While this approach is indeed appealing, it has still some drawbacks, e.g:
- generally, you can't tell what your topoligy your network will end up having, so forget about architecting one
- it does not guarantee that all your nodes will end up being networked within a fixed number of attempts (see the fig. 3 in the paper)
- it tends to require significant redundancy of interchangeable nodes to function well

Such approach can work well, say, for military field communications, but would be clearly suboptimal for building a corporate network.

And of course, as most of agent research, this is still too far from established technology ready for production.