Re: [SED-ML-discuss] Configuring algorithm-specific parameters

[Jonathan C. <jon...@cs...>]

On 13/09/2011 04:36, David Nickerson wrote:
> On Mon, Sep 5, 2011 at 9:17 PM, Richard Adams<ric...@ed...>  wrote:
>> Hello all,
>>   The new release of the Kisao ontology contains some information about
>> algorithm-specific  configuration parameters (e.g., error control
>> parameters).
>> Currently in SED-ML only general settings ( start time, end time, number of
>> Intervals )  can be defined, and an algorithm type.
>>   I'd like to open a discussion that we extend the 'Algorithm' element to
>> include these specific configuration parameters.
>>   For example the XML might look something like this:
>>     <algorithm kisaoID="KISAO_0000071">
>>        <algorithm_parameter kisaoID="KISAO_0000211" value="0.000001"/>
>>    </algorithm>
>> would set the absolute tolerance  for an LSODA algorithm. The 'value'
>> attribute could be a string, since the actual data type of the parameter
>> value
>>   can be resolved  from the Kisao ontology description for that parameter.
>>   Cardinality would be 0*, i.e., algorithm_parameter elements
>>   would be optional.
> That seems like a sensible approach, and if I remember the discussions
> correctly its also why algorithm was moved to be an element instead of
> an attribute. I should probably look myself, but does anyone know off
> the top of their head whether KiSAO defines default values for all the
> parameters for all the algorithms?

I can't see any defaults from a quick browse, and I'd guess they 
wouldn't be included (different implementations of the same algorithm 
are likely to use different defaults), but I'm not that familiar with KiSAO.

I also like the approach, although I think the element name 
"algorithmParameter" would be more consistent with the rest of SED-ML.

Best wishes,
Jonathan