how to address errno

Using the Multiple Run Fitter, praxis, etc..
Post Reply
loremauro
Posts: 3
Joined: Mon Nov 14, 2022 11:09 am

how to address errno

Post by loremauro »

Hi, I'm trying to optimize a Nav channel model so I'm taking ranges for the parameters that I want to optimize and giving them to an algorithm that performs the simulations and so on... my question is about how to handle a particular evenience:

exp(3099.23) out of range, returning exp(700)

this error is a result of the rate calculation done in the mod file (thus the parameters passed to the algorithms are disproportionate) and I was wondering, is there a way to address this error? Because it's not explicit like, for example, a ZeroDivisionError, which I can address directly.

Thanks
ted
Site Admin
Posts: 6353
Joined: Wed May 18, 2005 4:50 pm
Location: Yale University School of Medicine
Contact:

Re: how to address errno

Post by ted »

The first thing to check is the NMODL source code. "For what?" you might ask. The list of possible problems includes typographical errors (e.g. using a minus sign instead of a plus), failure to include scale factors needed to resolve inconsistencies of units, algebraic formulas that are written in ways that are prone to generate numerical overflow--but there are many others. Discovering and fixing some of these problems can be very difficult. If you like, I'd be glad to look at this file and tell you what I find.

"Why not simply make the optimizer limit the range of possible values of some parameter?" Fine, as long as (1) you're quite certain that the NMODL code is OK, and (2) you know what parameter is causing the problem, and (3) you have an idea of the reasonable range of values for that parameter, and can justify this to a reviewer (for example, concentrations and reaction rates should never be < 0).
loremauro
Posts: 3
Joined: Mon Nov 14, 2022 11:09 am

Re: how to address errno

Post by loremauro »

Thank you for all these guidelines, I'll e-mail you the mod file!
Post Reply