Re: Basic References on ABM validation/verification

[Ralf Stephan]

D4: 
> In general, I use two terms: "verification" for demonstrating the the model
> is (relativly) bug free, and "validation" for comparing the model results to
> some external data of mechanism.  For the two terms combined I tend to use
> "black magic" [grin].  Unfortunatly, the reality of validation is that it as
> complex as the systems we study.  Best of luck!

For example, my background is being a researching patient, I've programmed
as diverse things as speech recognition --- at the same time teaching the
use of better instances of it to disabled buddies ---, taking statistics
from interviewed studies and using machine learning to get more statistics,
optimizing a small industry process with electrical wheelchairs etc. 

So, complexity is always an issue and I'd say it is possible to start
with reduced data (the ancient five sources), build software on it,
test with less limited data and so on --- testing culture is widespread
in Java and Sather.

If you want to suggest The Meaning[tm] of 'bug' being equivalent to
The Meaning[tm] of 'comparing model results to some external data of[sic!]
mechanism' and finding nonisomorphisms, then writing a paper that shows
isomorphism between the physical/biological mechanism and a, say, Swarm
model, might also show the correctness of the code.  But what use is
the paper, then, if it doesn't include said hotly debated thing?


ralf
-- 
http://ME.IN-berlin.de/~rws/


                  ==================================
   Swarm-Modelling is for discussion of Simulation and Modelling techniques
   esp. using Swarm.  For list administration needs (esp. [un]subscribing),
   please send a message to <address@hidden> with "help" in the
   body of the message.
                  ==================================