On Sun, 15 Oct 2000, Sampo A Syreeni wrote:
On Sat, 14 Oct 2000, Jim Choate wrote:
On a similar vein, just about every somehow understandable version of free market theory is based on the assumption of a steady state market.
Is there a distinction between 'equilibrium' and 'steady state' in this context?
None. But I think both terms are easily misunderstood - in most cases we can analyze the local time behavior of differential equations based on steady state assumptions.
Good. I agree with your comment. The reason that I was asking was situations like 'A-B Chemical Oscillators' and B-Z generators from chemistry as well. There is a clear stability but there isn't a singular steady-state. It seems to me that a statistical mechanical market modelled as a CA could have similar structures (e.g. Glider Guns). It would be interesting to understand what that meant in real world behaviour.
Many of the folks I deal with have turned to cellular models. They seem to behave in a reasonable manner and they're reasonably easy to manage. My personal interest is in the application of modelling the near term politico-military situation with the two China's.
Cellular automata are nice and well understood. But isn't this quite a leap from analytic treatment to the direction of numerical simulation? Not that there's anything wrong with that, but to me some of the theoretical questions (like the question of fundamental stability) really seem to live more in the domain of equations.
Since it's well understood that CA's represent a general computing mechanism it's not a leap at all really. Just a different manner of instantiating them. I think the clear superiority comes from CA's treating a distinctly modular market modularly where'as DE's tend to consider a continous domain. people, time, items, services, and specie come in modular forms. CA's inherently instantiate the concept of 'I' in the model, allow each cell to have unique 'neighborhood' rulesets. The hard part is to figure out how to take that equation and convert it into a 'neighborhood' ruleset. The work of Wolfram and Rucker has been quite helpful in this aspect. Other nice features of CA's is that they are naturals for parallel processing. They also fit into OO programming concepts easily. When you couple this with concepts like 'small networks', 'cake cutting algorithms', evolutionary/genetic algorithms, and fuzzy algebra/logic, and include game theory in the 'neighborhood' ruleset you end up with a very flexible tool. This approach is very multi-disciplinary and differs in characteristics from normal CA's with simple 'survival neighborhood' rulesets. One difference that drasticlly changes the behaviour is the allowance of 'action at a distance' with respect to 'neighborhood' selection. Make the state of the current cell a function of distant cells and not just immediate neighbors. Along this same vein, instead of the state dependencies being immediate add 'speed of light' limits via diffusion mechanisms. You can also use things like 'Annealling Theory' and spin-glasses. As Rucker says, There is a better way. YOU can do it. Seek the Gnarl! ____________________________________________________________________ He is able who thinks he is able. Buddha The Armadillo Group ,::////;::-. James Choate Austin, Tx /:'///// ``::>/|/ ravage@ssz.com www.ssz.com .', |||| `/( e\ 512-451-7087 -====~~mm-'`-```-mm --'- --------------------------------------------------------------------