> Would it be helpful to encode this information into the model itself, or
> semantically, do you want to keep them distinct?  I suppose it's edging
> into the SEDML territory of 'what you do with a model' instead of 'what
> the model is'--I can imagine models where you vary the components of the
> strain tensor in one simulation to determine some other value, and in your
> second simulation you vary the other value to determine the strain tensor.

exactly. Furthermore, with CellML 1.1 and modular reuseable models,
you can easily end up with the same variable in a specific component
being an independent or dependent variable depending on how that
component is imported and instantiated into a higher level model.

CellML just gives you a way to encode the mathematics.
Simulation/analysis tools would typically have to analyse the math in
order to work out what needs to be computed, or the user needs to
provide further information to let the tool know what to do with the
math. This is definitely starting to get more towards SED-ML. In the
CellML simulation metadata we explicitly defined the independent
variable for a given simulation (only one was allowed), whereas this
information in not in SED-ML L1V1.

>> > My second question is a bit simpler: ?is there a way to define a delay
>> > equation in CellML and if so, how do you do it? ?(A delay being
>> > something like 'x, 2 seconds ago'.)
>> there is, but its not really used by any real models. The plan is to
>> make things easier in CellML 1.2, but the focus to date has been on
>> the use of instantaneous delays in event descriptions. You can see
>> https://tracker.physiomeproject.org/show_bug.cgi?id=70 and
>> https://tracker.physiomeproject.org/show_bug.cgi?id=1543 for some
>> comments along these lines, and hopefully Andrew or Randall will reply
>> with a better answer for you.
> What about tool support?  Is there support in any/some tools for delays?
> I ask from the perspective of a translator, converting models with delays
> from other systems to CellML.

None that I know of. There is some limited support in the CellML API
in terms of the code generation and possibly integration service, but
I'm not sure if that work has been shown to be correct for "real"
models. I seem to recall Andrew and/or Randall had some toy test
models that were getting correct behaviour. But probably best to wait
for Andrew or Randall to comment on this.

cellml-discussion mailing list

Reply via email to