Re: What are the advantages / disadvantages of Nueral Networks vs. ComplexAdaptive-Agent based modelling?

[jmerelo]

HI

address@hidden wrote:

 Everyone,I'm doing the programming for a project that attempts to simulate "how people choose" in order to simulate a small economy with lots of individuals, a couple of businesses, and a government that regulates.  I've read most of the Sugarscape book and have a limited but good-enough understanding of the complex adaptive-agent technique.How does this technique compare to nueral networks?  My impression of nueral networks is that they're much more complicated.  Does it even make sense to incorporate nueral networks into agent-based simulations?  I'm not even sure what they use Nueral Nets to do, anyway.

Of course it makes a lot of sense, and it´s been done already. Check out http://kal-el.ugr.es/pitis.html , and http://kal-el.ugr.es/%7ejmerelo/research.html , the paper on Accouting, management and Information Techologies.

In those papers, there´s an agent-based world in which each agent has got a neural net. In this case, besides evolving behaviors, agents can change their opinion during their lifetime. In the second article, this is used to simulate a marketplace with consumers, several products and publicity; agents can actually change their buying patterns depending on publicity. Which would be imposible if you used only agent evolution, as in SugarScape

WRT to what Glen said, if you want to deal with a general neural net, maybe it would make sense to implement it as a Swarm, but with the most usual neural nets, like backprop or Kohonen´s SOM, it´s much easier to just have them as objects, as it´s done in NeuroLib.

Good luck, anyways. I would be interested in hearing about any progress you make

J