Collective Intelligence in Social Insects

David Gordon's contest entry has just been published. It was awarded the Jury's prize.

Biological systems have long been a source of inspiration for AI. The latest paradigm to make the virtual leap comes from social insects - ant colonies and swarms. This article gives an overview of promising technologies to emerge from this field, looking at reasons 'swarm thinking' might come to characterise the new millenium.

http://ai-depot.com/Essay/SocialInsects.html

Nice essay! Very well written, and chock full of good information. I like the demo, too! Thanks for the contribution.

-Kirk

Excellent essay. One of the best I've read here. Well written and interesting. I sure hope to see more essays like this here.

Yes. I like this animal story. Great David Gordon.

In nature we see often a mechanism between different animals that also works within other more complex animals. It's a kind of fractal-like phenomenon, you can understand when you read the chaos/order view, also described here. I don't know if the AI warehouse has a keyword search tool.

It's then not strange in our brains we recognize the same kind of mechanisms.

Nature is a rich source. Don't always grab at a mathematical representation but describe in physical terms what is happening.

I know that this pre-physical approach is much more productive than the mathematical approach. I've copied your story and will study it more careful.

Thanks,

Ed van der Meulen

I wonder if you're aware of Tuevo Kohonen's work on self-organizing memories? In particular, his learning vector quantizer (LVQ) algorithms bear an interesting resemblance to the simple PSO algorithm, though they're solving somewhat different problems. In Kohonen's classifiers, the 'exemplar' nodes are pulled around by the data nodes in n-dimensional space in order to create the most efficient (quasi-Bayesian) classification scheme. There's a nice demo of the LVQ3 algorithem at http://www.neuro.sfc.keio.ac.jp/~masato/jv/lvq3/ (though the rest of the page is in Japanese).

Yes Alchemist, that's very interesting and useful. I know many applications of it, for instance searching of cancer cells in breasts.

A problem is it's time consuming or needing large computers so relatively expensive. We are searching for cheaper solutions.

But this mathematical solution is till now the best. I prefer other more fast working solutions, which are developed as well.

n-dimensional projections are used in more applications such as in PCA and ICA. These techniques reduce the info dramatically. Also for environment recognizing. But also here we look for faster solutions.

Thanks for you positive comment

Ed van der Meulen

hey all,
im doin a project on swarm intelligence and have started to study particle swarm optimization, i understand the concept and all and the theory behind, but i need a proper explanation of the algorithm, i mean i understand that the particles are effected both by their personal best position and the globlal best..but the equation stated, how has that been devised..i would be extrememly grateful..its this :
for each particle p
for each coordinate c
p.c = p.c + 2 * rand() * pbest.c - p.c + 2 * rand() * gbest.c - p.c

the line of the inner for loop, just need an explanation for it.

Hey ppl....m currently assigned a project on PSO...i have collected the required basic material bt have not been able to know or find out exact working and the pattern behaviour...can any one pls help me with that....coz m very much interested inmknwing the pattern logic...demo along with it would also be of gr8 help...u can mail me if any related information on cric85@yahoo.com thanks!!