Re: [SOSlib-discuss] Problem on simulating models with triggers andevents[Scanned]

[Rainer M. <ra...@tb...>]

Hi Eva,

Eric's hints are correct. SOSlib can handle SBML events, it is however 
not fully SBML-conformant. Here some more background:

SOSlib does NOT detect the exact time an event trigger fires. This would 
in principle be possible using CVODES' root finding capabilities - which 
however would also significantly slow down the solver. We plan to 
implement root finding first for the new SBML 'constraints', and later 
extend this to implement exact event trigger detection. However, I can't 
give a time-frame for that development.


setHaltOnEvent option:

Currently, setting the option setHaltOnEvent to 0 (-n in the 
command-line tool) activates trigger firing and execution of event 
assignments between the requested solver steps (--printstep on 
command-line, and CvodeSettings_setPrintstep(set, int i).

For many event models and purposes this is enough, but the 
user/application has to set printsteps accordingly to obtain the desired 
accuracy and to avoid missing of events. For example, SOSlib solves the 
event models in the official (but outdated) semantic-testsuite correctly 
when the appropriate printstep is chosen. Also I am not aware of any 
model in Biomodels where SOSlib event handling would cause real problems 
(but maybe Lukas can give some examples, if he's reading :) ).

The default for this option is to STOP the solver when events are 
triggered, as we can't guarantee the results. Thus the user/application 
is required to specifically activate it.


setResetCvodeOnEvent option:

When an event assignment resets variables Y of the ODE system f(Y,p,t) = 
dY/dt, then the CVODES solver must be reinitialized in any case, and 
this is always done. If the event assignment only changes the 
right-hand-side of the ODE system, it is recommended to also 
reinitialize the solver, as such discontinuities screw up CVODES' error 
estimations (as far as i understand it). Thus, it is recommended to 
activate the setResetCvodeOnEvent option, but for simple cases this is 
also not necessary - the solver is faster if not reset on such event 
assignments.
For the next release, also this option will be active on default!


new options in next release:

For the immediate next release, it is planned to introduce a simple new 
option, 'timeStepResolution', which activates internal integration steps 
between the then only external printstep. This will allow a finer 
resolution of event trigger detection without having all the values in 
the output. So e.g. in a simulation to time 1000 seconds, with printstep 
50, the event triggers are only checked every 20 seconds. You would need 
  printstep 1000 to detect at every second, but you would also get 1000 
result lines. With the new option one will be able to e.g. set 
StepResolution to 20 and Printstep to 50 to get the same result, but 
with only 50 lines of results.

If your application uses the IntegratorInstance_integrateOneStep and 
reads results between steps itself (rather then using storage of results 
with option StoreResults), you can easily implement higher detection 
resolution without overloading the output yourself.


And one more thing to note:

The new lack of (assignment) rule ordering restrictions in L2V3 would 
make it necessary to also account for correct order of rule evaluation 
in event trigger and assignment handling (see my recent email and 
responses to the sbml-discuss list). I hope, but I am not sure if I have 
time, to implement the possibility of such complex dependency trees (or 
DAGs actually) of event triggers, event assignments, and normal 
assignment rules also for the next release.

Please keep asking, if you have any more questions or suggestions on 
SOSlib's event handling.

Rainer




Eric Fernandez wrote:
> Dear Eva,
>  
> I am coming back to you. I have checked the options you need to activate 
> to enable the event evaluation during an integration. You must use:
> - cvodeSettings_setHaltOnEvent(cvodeSettings_t *set, int i) -> i=0 
> (default is 1) to continue after an event, otherwise it halts, which is 
> the default!
> - cvodeSettings_setResetCvodeOnEvent(cvodeSettings_t *set, int i) -> i=1 
> recommended, free and restart integrator after an event.
>  
> Hope this helps,
> Eric
> 
>     ------------------------------------------------------------------------
>     *From:* Eva Sciacca [mailto:sc...@mi...]
>     *Sent:* 17 October 2008 20:37
>     *To:* sbm...@li...
>     *Subject:* [SOSlib-discuss] Problem on simulating models with
>     triggers andevents[Scanned]
> 
>     Dear all,
> 
>     While I was trying to do some testing using SOSlib I had this errors
>     on running one model from BioModels DataBase.
> 
>     Warning    10001    The model contains events. The SBML_odeSolver
>     implementation of events is not fully SBML conformant.
>     Results will depend on the simulation duration and the number of
>     output steps.
>     Error    20001    Event Trigger 1 (leq(UbE, 0.1)) fired at time 16.
>     Aborting simulation.
> 
>     So, the current version 1.6.0 does not support the triggers? What
>     about the newer version? Are you going to solve this problem? When
>     it will be available???
> 
>     Eva.
> 
> 
> ------------------------------------------------------------------------
> 
> -------------------------------------------------------------------------
> This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
> Build the coolest Linux based applications with Moblin SDK & win great prizes
> Grand prize is a trip for two to an Open Source event anywhere in the world
> http://moblin-contest.org/redirect.php?banner_id=100&url=/
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> sbmlsolver-discuss mailing list
> sbm...@li...
> https://lists.sourceforge.net/lists/listinfo/sbmlsolver-discuss