|
|
|
|
Subject: Re: some questions
From: Nicolas Becu (nicolas_becu@yahoo.com)
Date: mer jan 09 2002 - 16:32:40 CET
Hello Niek,
and happy new year too
I'v put my comments in you original text
----- Original Message -----
>>X-Incognito-SN: 668
>>Date: Wed, 09 Jan 2002 12:53:02 +0100
>>From: Niek.Koning@Alg.AAE.WAU.NL
>>Subject: some questions
>>To: cormas@cines.fr
>>Cc: Wiebe.Aans@Sg.OSA.WAU.NL, Annemarie.vanPaassen@Alg.VLK.WAU.NL
>>Reply-To: Niek.Koning@Alg.AAE.WAU.NL
>>X-Incognito-Version: 5.0.1.93
>>X-Loop: cormas@hermes.cines.fr
>>X-Sequence: 327
>>
>>Happy new year,
>>
>>I have worked further on the IronAge model that Wiebe Aans, Annemarie van
>>Paassen and I constructed during the MAS-course in Wageningen. I now have a
>>version which shows some interesting behaviour.
>>
>>I now have a few questions:
>>
>>1. When running an N-times simulation (with 'step: t nTimes: n') I only get
>>graphs for the last run. I suppose something must be changed in the
>>updateCharts command, but what is it?
I think you should include in your step, an export method of your charts that activates at the end of each simulation run.
You'll have to export into a file that include the N value of the nTimes method, so that the exported files do not save on each other.
You can find some export methods that will help you in the class of the Cormas object.
Otherwise I've wrote some methods that allow to export your graphs in csv file that can be read with Excel (tell me if you need)
>>
>>2. How can I systematically explore the model's behaviour for different
>>parameter values?
>>
That points up a question that I wanted to rise on the Cormas list. What your talking about is some kind of tool that would help us to make our simulation projects easier (a set of simulations with different parameter values). And I think that we're quiet a lot to need such tool.
Well from what I know, for the moment you have to do it manualy. There are no generic tool that you can file in and use directly. Some VW modellers, I think, have done such kind of tool with an appropriate interface etc... but they can't be used directly on other models.
That's why I would suggest two things :
A- couldn't we buit an interface that would be added to Cormas 2002 , that would be dedicated to simulation projects.
Of course to do so, we will have to buit our models in order that this interface will be able to recognise, collect and change the parameters of our models. But with a bit of thinking, I think it's quiet easy and feasable.
The interface would be something like : 1/ choose the parameters you wan't to modify ; 2/ enter the values ; 3/ how many time do you want to run the simulation; 4/ name of the file you want to export to ; etc...
B- the second think I'll like to suggest is, instead of having some additional tool of Cormas all around, why not add a button on the Cormas interface that would facilitate the connexion of your model to the tool you want to use. (ex. on the picture bellow). Then, we have to agree that, when one builts a additional tool, he makes it under certain conditions that will allow Cormas to recognise it as a "Cormas Tool". So another user, can file in the tool, import it with Cormas, and then call the tool and open it through the additional Cormas button.
I guess that, if we're all interesed by such an architecture, the Cormas team will be pleased to help us define an appropriate architecture for the Comas Tools. Isn't it ?
>>3. The model contains various stochastic elements and different simulation
>>runs give quite different results. Nevertheless, there are certain recurring
>>patterns, depending on parameter values. But which features of the model
>>exactly lead to these patterns is not so easy to see. What is the best way to
>>examine this?
I'm not shure to understand this one. Are you talking of spatial patterns on the Cormas grid ?
If yes, then I would suggest to find a way to characterize those spatial patterns automaticaly. When done,then you run different scenarios with different parameter values and you identify at which step of your set of simulations the pattern parameter is changing.
>>
>>Greetings,
>>Niek Koning
See you
Nicolas
_________________________________________________________
Do You Yahoo!?
Get your free @yahoo.com address at http://mail.yahoo.com