Small World Networks

classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|

Small World Networks

Taymaz Khatami

Hi everybody

 

I have a huge problem in using Repast’s SmallWorld method. Basically it is a scaling problem. When creating networks with 250 Nodes everything works fine. But I need to run my simulations with 1000 agents. The algorithm provided by Repast takes to long (approximately 1 day) to create the Network and rewiring the edges with a certain probability.

 

My Question: Is Repast supposed to simulate such large Networks? Or do I just need more CPU Power (currently: Intel Xeon 3.6 GHz HT)?

I wonder if the error is within my own code. But each run takes a few seconds to iterate through 100 Periods and as mentioned above several hours to create the Network (createWattsStrogatzNetwork) before running the simulation.

 

Thx in advance.  

Reply | Threaded
Open this post in threaded view
|

RE: Small World Networks

rnick
Can you email the bit of code you use to create the network -- e.g. createWattsStrogatz(...). I just did a few quick tests and a few variations on:
 
List net = NetworkFactory.createWattsStrogatzNetwork(1000, 1, .2, DefaultNode.class, DefaultEdge.class);
 
take less than a second.
 
thanks,
 
Nick
-----Original Message-----
From: [hidden email] [mailto:[hidden email]]On Behalf Of Taymaz Khatami
Sent: Friday, September 09, 2005 3:49 AM
To: [hidden email]
Subject: [Repast-interest] Small World Networks

Hi everybody

 

I have a huge problem in using Repast’s SmallWorld method. Basically it is a scaling problem. When creating networks with 250 Nodes everything works fine. But I need to run my simulations with 1000 agents. The algorithm provided by Repast takes to long (approximately 1 day) to create the Network and rewiring the edges with a certain probability.

 

My Question: Is Repast supposed to simulate such large Networks? Or do I just need more CPU Power (currently: Intel Xeon 3.6 GHz HT)?

I wonder if the error is within my own code. But each run takes a few seconds to iterate through 100 Periods and as mentioned above several hours to create the Network (createWattsStrogatzNetwork) before running the simulation.

 

Thx in advance.  

Reply | Threaded
Open this post in threaded view
|

RE: Small World Networks

Meifeng Luo
In reply to this post by Taymaz Khatami

Is this a bug in somewhere in the RepastOMLayer.prepare? It often occurs when I execute the simulation model, resulting in the disruption of displaying simulation result. The error code is like this:

 

java.util.ConcurrentModificationException

          at java.util.AbstractList$Itr.checkForComodification(AbstractList.java:449)

          at java.util.AbstractList$Itr.next(AbstractList.java:420)

          at anl.repast.gis.display.RepastOMLayer.prepare(RepastOMLayer.java:212)

          at com.bbn.openmap.layer.policy.StandardRenderPolicy.prepare(StandardRenderPolicy.java:68)

          at com.bbn.openmap.layer.OMGraphicHandlerLayer$LayerWorker.construct(OMGraphicHandlerLayer.java:626)

          at com.bbn.openmap.util.SwingWorker$2.run(SwingWorker.java:109)

          at java.lang.Thread.run(Thread.java:595)

 

Thanks!

 

Meifeng Luo  

Reply | Threaded
Open this post in threaded view
|

Re: Small World Networks

Björn Lijnema
In reply to this post by Taymaz Khatami
I think there might be a bug somewhere in your code. I've been working
with RePast generating networks on a computer slower than yours and
generating networks has never taken any significant amount of time. As
Nick Collier said, less than a second should be the amount of time
needed.

I know that visualizing networks takes _a_lot_ of time, once you get
past a few hundred nodes. So if you are displaying the network, that
can at least be a contributing factor.

But, I'd check for anything in the code around where you generate the
network and see if there is something that slows you down there. There
might be some accidental loop or similar. You shouldn't need more CPU
power.


Björn Lijnema


-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
_______________________________________________
Repast-interest mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/repast-interest
Reply | Threaded
Open this post in threaded view
|

Re: Small World Networks

jerry.vos@gmail.com
In reply to this post by Taymaz Khatami
One place you may want to look in particular is at the construction code
for your node and edge classes. The create methods use reflection to
build your nodes and edges using the default [ClassName()] constructor,
so I'd check through what happens when you create an edge/node with that
constructor and see if any wierd behavior is going on.

Jerry

----- Original Message -----
*From:* "Taymaz Khatami" <[hidden email]>
*To:* [hidden email]
*Sent:* Friday, September 09, 2005 2:49:22 AM
*Subject:* [Repast-interest] Small World Networks


> Hi everybody
>
> I have a huge problem in using Repast’s SmallWorld method. Basically
> it is a scaling problem. When creating networks with 250 Nodes
> everything works fine. But I need to run my simulations with 1000
> agents. The algorithm provided by Repast takes to long (approximately
> 1 day) to create the Network and rewiring the edges with a certain
> probability.
>
> My Question: Is Repast supposed to simulate such large Networks? Or do
> I just need more CPU Power (currently: Intel Xeon 3.6 GHz HT)?
>
> I wonder if the error is within my own code. But each run takes a few
> seconds to iterate through 100 Periods and as mentioned above several
> hours to create the Network (createWattsStrogatzNetwork) before
> running the simulation.
>
> Thx in advance.
>


-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
_______________________________________________
Repast-interest mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/repast-interest