RE: large-n simulations

[Darold Higa]

Oops, big correction there...I meant to say 400-500 iterations, with an end
population of around 1500-2000 agents.  I have run it in a few hours if I
turn off saving the displays to the hard drive.  I can speed it up even more
if I turn off the nonessential graphical elements.

Swarm is great for doing large-n agent-based models.  I have hopes to be
able to run simulations with 5000+ thinking agents interacting in a market
and in a physical environment in the next stage of my simulation.  That is
probably at least a year or two off, so I hope that I will be running it on
a 4GHz machine by then.

My biggest difficulty with large-n simulations was memory.  I wanted to give
every agent a larger chunk of memory, but I began to have problems managing
some of it (more a limit of my programming ability than Swarm's fault).

I wouldn't be surprised to see social science simulations in the 10,000+
agent range in the near future.  Is anyone already doing it?

Darold Higa
University of Southern California
School of International Relations


                  ==================================
   Swarm-Support is for discussion of the technical details of the day
   to day usage of Swarm.  For list administration needs (esp.
   [un]subscribing), please send a message to <address@hidden>
   with "help" in the body of the message.