Exceptions in Parameter Estimation

13 views
Skip to first unread message

Luis Andrés Rojas Matey

unread,
Aug 16, 2021, 3:16:12 PM8/16/21
to COPASI User Forum
Hi!

When I run a Parameter Estimation, by using NL2SOL method with iteration limit of 2000, I get a popup up with exceptions (see the attached image).

I tried to reduced the boundaries (lower and upper) and also tried to check which species produced the exceptions by doing Time Course simulations, but nothing works for me.

I have several questions:
  1. These exceptions act as "warnings", right? (since I noticed there are several in the popup but the final report is created)

  2. Is there an easy way to check which species (and which parameters) are causing the exceptions? (since Time Course simulations show NO negative values, but also NO values that "grow extremely fast")

  3. How can avoid the exceptions? (different than changing the boundaries)

Thanks in advance.

Regards.


Exception.png

Hoops, Stefan (sh9cq)

unread,
Aug 16, 2021, 4:01:37 PM8/16/21
to copasi-u...@googlegroups.com
Hello Luis,

You should also have has a warning that a number of evaluations failed.
This is an expected result especially if you use steady-state
experiments (or checked start in steady-state) and should not worry you
if you are OK with the final result. This information is provided by
COPASI so that a user has all information necessary to fix an eventual
problem.

Thanks,
Stefan
--
Stefan Hoops, Ph.D.
Research Associate Professor
Biocomplexity Institute & Initiative
University of Virginia
995 Research Park Boulevard
Charlottesville, VA 22911

Phone: +1 540 570 1301
Email: sho...@virginia.edu

Frank Bergmann

unread,
Aug 17, 2021, 4:03:52 AM8/17/21
to COPASI User Forum
Hello Luis, 

as Stefan mentioned, yes indeed those are warnings that some iterations failed due to NaNs encountered during the estimation. There is not an easy way to check which ODE was responsible for this. What you can do, is to look at the 

Tools\Check model 

feature, to see if maybe there are some species that would be consumed even if they are 0. Other than that you would have to look at your equations and look out for divisions by zero, or logarithms of zero or negative numbers. And modify them slightly to avoid those. 

If that does not help, you could send us (or me) the model and let us have a look at it. 

best
Frank

Reply all
Reply to author
Forward
0 new messages