Forums

Local parameters seem wonderful until you find that you must refer to them
from outside their scope. Optimization is a case in point.

Optimization requires a vector of unique parameter names/ids, and so to fit
an entire model to data requires that global parameters be passed to the
optimizer. I have to agree that it's very convenient to be able to use Km
again and again locally, but as soon as you begin a parameter optimization
procedure I believe you are going to have to give your parameters global
names, or at least global ids, anyway. So, like Herbert, I have a strong
preference for global parameters.

The trouble is that it's essentially impossible to make a parameter
permanently global. Within a given model, of course, this is easy. But as
soon as you decide to combine two models to make one larger model, all your
"global" parameters suddenly become embarrassingly local.

So from my perspective, local and global parameters both have their uses.
Furthermore, any attempt to legislate one or the other out of existence is
doomed to fail on logical grounds. No software tool can afford to treat the
global/local distinction lightly because there are important use cases where
global parameters become local and others where local parameters become
global.

My two cents,

Robert D Phair, PhD
Chief Scientific Officer
Integrative BioInformatics Inc
Rockville, MD 20854 USA
301.315.8114

-----Original Message-----
From: sbml-discuss-bounces@caltech.edu
[mailto:sbml-discuss-bounces@caltech.edu] On Behalf Of Herbert Sauro
Sent: Thursday, July 21, 2005 4:08 PM
To: SBML Discussion List
Subject: RE: [sbml-discuss] Proposal: simple mechanism for referencing
ratelaws



-----Original Message-----
From: Nicolas Le Novère [mailto:lenov@ebi.ac.uk]
Sent: Thursday, July 21, 2005 6:51 AM
To: SBML Discussion List
Subject: RE: [sbml-discuss] Proposal: simple mechanism for referencing rate
laws


>Oh, my. Could-it be the end of the dreadful "global" parameters?

Odd you should say that because I am in the process of moving completely
over to global parameters because local parameters cause me so much trouble!

Herbert Sauro

--
Nicolas LE NOVÈRE, Computational Neurobiology, EMBL-EBI, Wellcome-Trust
Genome Campus, Hinxton, Cambridge, CB10 1SD, UK
Tel: +44(0)1223 494 521, Fax: +44(0)1223 494 468, Mob: +33(0)689218676
http://www.ebi.ac.uk/~lenov