Sam Isaacson

Yes, that could use some rewriting for clarity. Thanks for pointing it out.

rng management and seeding is a bit of a mess in general. Here we pass `rng`s via the problems and store them in the aggregators, StochasticDiffEq doesn't allow setting an...

@ChrisRackauckas would creating a SciML rng interface that works across serial/parallelism modes make sense as a small grant project? Or is it too complicated to expect someone to tackle in...

To make passing via `solve` work we would need all integrators in OrdinaryDiffEq and StochasticDiffEq to store the `rng` and have a standard way to get it from them (i.e....

A problem with that approach, i.e. user generated per thread/process seeds, is that it isn’t how all (many?) parallel rngs are designed to work. For example, some generators handle setting...

Can you add some literature references showing the equations one wants to solve for some simple examples?

Is this a Catalyst issue or a general MTK issue with the new symbolic indexing interface changes? If the latter this should be an issue over there.

This seems to be the culprit: https://github.com/SciML/Catalyst.jl/blob/005a58fa70b8a5a85073a60c788cf55b45136ce7/src/dsl.jl#L788 We need to instead directly build the observable variables with the independent variables specified.

@TorkelE it would be good to fix this for V15 since it technically changes the semantics of these variables.

If a symbolic is created via an observable the generated symbolic stood be the exact same as if it was created with `@variables`.