[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
re: fest discussion -- model testing
From: |
Matthew Hare |
Subject: |
re: fest discussion -- model testing |
Date: |
Thu, 26 Feb 1998 17:31:06 GMT |
Although I am not able to go to swarmfest this year (alas) I would like to
'fourth' the call for a meeting on testing and verification, as a
mightily important subject to discuss.
I'd add to the set of re-implementation problems: re-scaling -
that of deciding whether or not models originally designed at one level
of the organizational hierarchy (e.g. mathematical population based ones)
have been accurately
re-implemented at another lower level (e.g.a-life individual based). The
problem I
believe is
exacerbated by the conjecture that the set of possible system responses that
can emerge from a model of the same system developed at a lower level will/can
be greater (and perhaps different) than
the set of system responses generated by a model developed at a higher level.
(This is
after all what I thought was the power & point of doing A_Life simulation).
Hence, simply checking whether or
not the re-implemented model mimics the behaviour of the original higher level
model may prove little either way. This is a problem if we are to persuade the
scientific
community in our own disciplines of the worth of changing modelling paradigms.
We must then take both models back into the real world and compare their
'predictive' powers. But to compare models we need to find a common language
of description for the models' performances. Can we stick to using mathematical
descriptions of model performance? .... a question to resolve, maybe...
No answers yet, but as some of you will know from last
year's Swarmfest, over here we're developing a model testing environment for
swarm apps in order to facilitate the testing/comparison and analysis of
model components and modelling assumptions. So Catherine, tell me more
about your ground zero approach and I'd love to hear how the swarmfest
discussion goes - Can interested parties continue the dialogue on the net
beyond the fest?
As a rather backseat member of the swarm community over the last year it'll
be a great pleasure to be back maybe for swarmfest III with some things to offer
back to you all (fingers crossed)!
have a good fest,
Matt
~~~~~~~~~~~~~~~~~~~
Land & Environmental Management Systems Team,
Macaulay Land Use Research Institute,
Aberdeen,
Scotland.
==================================
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.
==================================