Add more interfaces to the result of a PINN

[killah-t-cell]

Say I defined and solved a PDE with NeuralPDE and I have the result. I want to do things with that result beyond plotting it in a space. It would be great to have an interface that would let me access the solution of each variable in space.

Sort of like DiffEq.jl has sol.t[8] interface, it would be awesome if I could do res.y[3] to access the values of my solution.

[KirillZubov]

Only it is a continuous solution, not discrete, sort of like res.y(x) where x in domain_x

[KirillZubov]

but yeah, I agree, it will be cool to have some solution wrapper that can make base manipulation to a solution. something like:

sol.y(0.1)
sol.approx(dx)
plot(sol)
sol.res

[killah-t-cell]

Good point. Exactly. It would also make 3D + time plotting easier.

[ChrisRackauckas]

And I'm not sure this is the right level to be thinking about the solution. I think there needs to be a common interface around all consumers of the PDESystem interface for how to then represent and plot the PDE results. That said, since the results take many different forms, that's pretty hard.

[killah-t-cell]

https://github.com/SciML/SciMLBase.jl/issues/112 created an issue in SciML Base and paraphrased you.