Implement distribution support

[matthiaskoenig]

List of functions in addition of min, max, prod, sum:

• count** : returns number of all elements
• mean** : returns mean of all elements
• sdev** : standard deviation of all elements
• variance**: variance of all elements
• uniform() : random value in [0, 1); 0 included, 1 not included; gives a single scalar number
• uniform(min, max): random value in [min, max); min included, max not included; gives single scalar number (min/max both floats)
• normal(mean, stdev): random draw with given mean and stdev.

**: New 'reducing' functions: add to min/max/other list

[jonrkarr]

Is this intended to include reductions over any dimension (e.g., variable, time, space)?

On a related note, has there been discussion about matrix computations?

[jonrkarr]

I think it would also be good to have examples with integer and Boolean-valued results to illustrate how this can be used with discrete and logical simulations.

[luciansmith]

We seem to be settling on something a little different at this point, such that your proposal would change to:

• Add KiSAO terms for 'count' and 'variance'.
• Implement support for uniform and normal draws.

The latter is already in the spec--do you have an implementation for it? Do you need something to do so? Or at this point do you want to drop them?

[jonrkarr]

Lucian's proposal seems good to me

• Additional KiSAO terms for other aggregate methods. See SED-ML/KiSAO#93
• Number generation seems fine as outlined in the specs. For example, BioSimulators and 14 downstream tools support all of these.

[luciansmith]

All the aggregation functions are now in KiSAO, but not the draws from distributions. Moving this to be an L2 issue, though if we think we can move this to be a KiSAO issue, that's fine, too.