[Swarm-Modelling] Re: [Repast-interest] Distributed RePast

[Darren Schreiber]

Excuse the crosspost, but I think the discussion is useful to both  
communities.  Please post followups on the RePast list to avoid mess.


I have been thinking along the same lines as this for a little while.   
UCLA's brain mapping group is moving towards a Grid setup and I think  
the technologies are going to be getting to the place where this  
shouldn't be too hard to accomplish.  They've already got this cool  
toolkit called pipeline that allows you to take data from one program  
into another really easily in an automated manner.  So I could imagine  
running simulations, then dumping the output into a stats package, and  
then dumping out the graphs all automatically.

Furthermore, I would contend that there are some very good reasons to  
go this route methodologically (in addition to the geek/cool factor).   
In a paper, I've written on model evaluation (aka validation) I argue  
for a multimodal approach to model evaluation.  One down side is that  
some of these modes require a lot of computing time.  Running the model  
to fit it to empirical data sets, exploring the parameter space,  
robustness testing, GA and GP motivated iterations of the model, etc.

My grand vision is a set of tools where you could pop your model in and  
the system would evaluate your model in a variety of conditions and for  
the purposes that you are trying to establish.  Since this is just  
rerunning the model again and again it is perfect for poor-man's  
parallelism (like Drone or some other type of scripting).  But it would  
be even better if it could be distributed.  Perhaps, all of us using  
RePast would someday be able to donate the spare cycles we have to  
model evaluation for others.

In my mind, this is the right way to approach the epistemological  
problems of agent-based modeling.  Formal models have their "proofs."   
And, stats has its 95% confidence interval.  The problem with ABMs  
(which applies just as much to formal models and stats when you think  
deeply about it) is that we can so easily vary our assumption set and  
the parameter space is so large that we  need a new epistemological  
foundation for truth claims.

In political science, Chris Achen has a nice paper where he argues for   
"Rule of Three" --- I should not buy your statistical model if you have  
more than three explanatory variables because the parameter space is  
too big to really understand the model.  This is exacerbated when we  
think of the proliferation of statistical models one can apply to a  
data set Ordinary Least Squares, Logit, Probit, Maximum Likelihood  
Estimation, etc...  Each of these is called for under different  
assumptions about the data and the problem, but when you can not only  
choose the EVs, but the model, and everything else, then it is easy to  
see that a 95% CI isn't as obvious as a it should be.  There needs to  
be a much broader regime for model evaluation.

My contention is that Grid computing or something like it is justified  
for ABMs because we are ever more in a world where we risk generating  
non-robust models that do not satisfy validity testing, but being blind  
to these problems since the epistemological standards developed in the  
early 20th Century have been satisfied.

        Darren



On Thursday, August 7, 2003, at 09:51  AM, Stephen C. Upton wrote:

> Max,
>
> I'd be interested in collaborating. For a project I've been working on  
> for the past several years, we do something similar using Condor  
> (http://www.cs.wisc.edu/condor/) as our distributed computing  
> mechanism, and it works with a couple of our simulations, 2 in java  
> and 1 in pascal.  Another idea might be to use the Globus stuff, all  
> in java, and set up to do grid computing. And a final idea is possibly  
> using something like JADE (http://jade.cselt.it/), which can activate  
> agents on different machines. Obviously, all require some learning,  
> and there are likely other possbilities, one you mention below. The  
> advantage of the above software is that most take care of registering  
> machines, handling security, and all of that other cr*p that's not  
> very interesting! ;-)
>
> steve
>
> Max J Cantor wrote:
>
> > Has anyone considered the creation of a address@hidden like screen saver  
> > for
> > RePast?   Nothing to complex, something that just does runs in the
> > background and uploads the results back to a central server.  I was  
> > just
> > comtemplating the idea because on college campuses like mine there are
> > tons of computers idling in labs.
> >
> > Off the top of my head, I see a server daemon that sends out the  
> > model jar
> > and batch files representing a single run of the main batch file to  
> > the
> > clients which run the model then upload the results.
> >
> >
> >
> > Given java RMI, and some other technologies, this should not be  
> > incredibly
> > hard to implement, so is there any interest in usage or collaboration?
> >
> > Max
> >
> >
> >
> >
> > -------------------------------------------------------
> > This SF.Net email sponsored by: Free pre-built ASP.NET sites including
> > Data Reports, E-commerce, Portals, and Forums are available now.
> > Download today and enter to win an XBOX or Visual Studio .NET.
> > http://aspnet.click-url.com/go/psa00100003ave/ 
> > direct;at.aspnet_072303_01/01
> > _______________________________________________
> > Repast-interest mailing list
> > address@hidden
> > https://lists.sourceforge.net/lists/listinfo/repast-interest
>
>
>
>
> -------------------------------------------------------
> This SF.Net email sponsored by: Free pre-built ASP.NET sites including
> Data Reports, E-commerce, Portals, and Forums are available now.
> Download today and enter to win an XBOX or Visual Studio .NET.
> http://aspnet.click-url.com/go/psa00100003ave/ 
> direct;at.aspnet_072303_01/01
> _______________________________________________
> Repast-interest mailing list
> address@hidden
> https://lists.sourceforge.net/lists/listinfo/repast-interest