David Widmann

Yes, as mentioned in the initial discussion in DelayDiffEq, IMO we should just use callbacks. The solver handles discontinuities fine, a common example in the tests is actually a discontinuity...

I'm sorry, I can't focus on this right now and I would have to look up the adjoint equations first. However, in my opinion, the major question when implementing the...

My point was that it is not possible to specify the Jacobians manually :shrug: Similar to the ODE case, there should be an interface that allows users to specify all...

> it shouldn't be a problem to add it It is trivial if there exists a reasonable design - it is something that has to be addressed in DiffEqBase in...

> We can make a stronger save_idxs for DDEs that makes it assume no history on the non-saved parts if needed. I think I already made some optimizations here. At...

> So this means we should be able to add the jacobians directly to the DDEFunction, right? Yes, exactly, the Jacobians and all other terms required in the sensitivity equations....

Indeed there is already an issue: https://github.com/SciML/OrdinaryDiffEq.jl/issues/335

I just had a closer look at your code (for the first time, I have to admit), and unfortunately I think the implementation only works in your special case (at...

Maybe the cleanest approach would be to add some special types of `DDEFunction`s that encode additional information (e.g., with only constant delays of order zero) such as certain mathematical structures...

The central point of the proposal above is to "pushforward" automatically the type of random samples from a lower level to the arbitrarily complex structure of the samples from a...